Of Hacks and "AI"
[blogger note]: I am rushing out this post, which likely means it won't even meet my nearly non-existant standards for publishing a piece, which is to attempt for a post to be at least coherent. I am doing so because I have become anxious that the "AI" bubble might collapse before I even get around to publishing this now dusty old draft. I want to be clear however - that I am by no means confident the bubble will burst any day now, or even soon. I can certainly see the possibility that both Anthropic and OpenAI might have successful IPO launches and stave off their creditors for a little while. There is also a sliver of a fraction of a percentage, that by some unanticipated means Generative LLMs find use cases that allow them to generate a profit. It just seems more likely than before as the media seem to be reporting on the fact that "AI" has these concerning features now of being a) unprofitable, b) generate no objective productivity gains for the productive businesses attempting to integrate them, c) are broadly unpopular, being that some plurality of people don't want to adopt them.
New Intro
As I write this the US is on the cusp of heading to the midterm elections, at least some sizeable minority of people who can vote that will, with what is now by polling data, the least popular President in US history of polling.
It is an interesting case study in our collective psychology, because driving the unpopularity of the US President is inflation/cost of living. The general price level relative to incomes. He was very likely permitted a second term in 2024 election, because of the high-but-decelerating inflation of the Biden administrations tenure, that was known to be attributable to COVID supply chain shocks (so supply side, rather than demand side, inflation) and Corporate profit gouging.
However, I would not say that it was generally known, it was however known by expert political economists like Mark Blythe (for example) a Professor at Brown University and public intellectual, and some economically literate lay-people like (for example) Andrew Yang, a former democratic presidential candidate and NYC mayoral candidate.
Crucially for this analogy, the people that did know, in the lead up to the 2024 Presidential election, made EXPLICITLY CLEAR that Trump's cornerstone campaign policy of introducing tariffs, would have inflationary effects. And that was just the Tariff policy. An economic expert need not and likely could not anticipate that even Trump would do anything as bone headed as go to war without congressional approval with Iran on the very day former president Bill Clinton was due to testify about his relationship with Jeffrey Epstein.
For a certainty, I HATE how brilliant a certain smug and tedious sort of left-wing youtuber thinks the "leopards ate my face" meme is, and it is probably unfair, smug and tedious in actuallity - it blames a victim of a concerted and coordinated campaign to deceive them for power - for falling for it.
The fact being that the Tories, the Republicans, Reform UK, the Coalition (AUS), One Nation etc. etc. don't call themselves "We will make the rich richer and the poor poorer" party. Or even the "Mortgaged futures" party. They predate on people's economic illiteracy, and to some extent their opposition does too, albeit less effectively.
Better is Maya Angelou's heuristic:
"When someone shows you who they are, believe them the first time."
And Maya being a poet-par-excellence, no doubt chose her words carefully she says "shows" not "tells" and on this ground, the right wing parties of the free democracies of the world have almost universally shown the popular vote who they are, they are not responsible managers of the economy (Raegan, Thatcher, Bush, Howard, Cameron, May, Johnson & especially Truss, Abbott, Morrison...I could go on but I'd have to start using google) They almost universally exist to relieve those in their nations most capable of paying, of the tax burden and instead borrowing from the rich to enrich the rich and impoverish the public, they also routinely attempt to redress the debt they put themselves in by foregoing revenue and selling off assets, by running austerity programs while facilitating an ever growing number of private sector monopolies.
This however is very complicated, it is no where near as obvious to comprehend as a party of leopards campaigning on the platform of eating people's faces. Neoliberalism survives Darwinism, on the same grounds that viruses and parasites that kill their hosts survive - by getting around our immune systems long enough to jump to another host.
I shouldn't have made that analogy because I am not generally biologically minded. The first thing we can do in this analogy is suppress an immune system by making sure enough people are economically illiterate, they simply have no defenses. We can also sneak by people's critical faculties by appealing to intuition and emotion.
For example - it is pretty easy to convince people that tax cuts are good for them because they will have more money. We have seen in recent years, the general public get temporary and quite restricted tax-relief, while the same legislation grants unrestricted and permanent tax relief to the wealthiest. So like, you give people a $20 food court voucher to get them to agree to vote that an Oil Tycoon should pay no estate tax when they pass on billions to the children lucky enough to be born to them.
It's also intuitive to argue that the government lives beyond its means, so it needs to cut spending and then quote to a person how much tax revenue is spent just paying the interest on the government debt.
Now, here is where the media is an unmitigated fucking failure at rendering a public service because instead of combatting these persuasive narratives with informed, factual counter narratives, they amplify them. They fucking amplify them, because the present day standard of journalism has a modal value of 'hack' complete fucking hacks.
The narrative is pretty simple - government spending is a huge part of the economy, to cut government spending is to basically induce a recession. Instead of cutting spending, the government could cut borrowing and increase revenue by taxing wealth instead of work.
The media could also explain, that public spending is also analogous to getting some kind of bulk-buy group deal - you get lower prices for goods and services if the buyers all form a negotiating block. Cutting taxes might seem like you are putting more money in your pocket, but instead of being like "well we'll all negotiate the price of healthcare together" we instead say "I'll just ask the doctor how much I owe him when my leg is broken and work it out between us."
Now, the media has been fucking failing the public hardcore in it's coverage of "AI" but again, just like Trump's tariffs causing inflation - if companies valued at 1-2 Trillion fall to (the floor's the limit but even) $100s of billions, I want to be clear - not only was an "AI" bubble bursting an entirely predictable risk, it was predicted. It was predicted almost as soon as what we now call "AI" went viral.
Now not everyone has the necessary discernment capacity to have seen with their own eyes, that "AI" was actually, not much of a gamechanger. But the evidence has been available to fucking everyone with eyes for a long time - that "AI" is more annoying than impactful.
Ed Zitron was recently interviewed by Jon Stewart on his "Weekly show" podcast, and many commenters noted Jon's particular question of "why can't somebody just ask AI how to make AI profitable?" as a kind of scalpel to cut through the bullshit of headlines like "AI just solved some 40 year old mathematical proof" pointing directly to the cognitive dissonance that "AI" can do all these amazing things allegedly, but nobody seems to have thought of using it to make its own development financially viable - something many a cafe startup entrepreneur figures out with pen and paper. (to be clear, many cafe owners figure out how to make their business viable, whereas "AI" "super-intelligence" cannot figure out how to train a LLM frontier model in a way that costs are proportional to what anyone is willing to pay for its limited use-cases)
In situations like this, where the media has literally driven most of the world mad, into not believing the evidence of their own eyes - I think of Kramer's "Oranges speech" and Bill Hick's CNN bit.
Again, to be clear, availability bias often puts us out of touch with reality, like Dan Gilbert's example of lotteries where we see news stories all the time about lottery winners, but if it was mandated that to interview a lottery winner you have to give equal time to every lottery loser -a single weekly draw would require 9.5 years to give a 30 second interview to every loser and that's with no sleep, no commercial breaks, just 9.5 years of solid 24 hour programming.
You may also have heard the 'absence of evidence is not evidence of absence' and skeptic Matt Dillahunty concurs but ammends 'an absence of evidence where evidence is expected should lower our confidence.'
And to believe in the power of "AIs" future, I would expect to see some major market shakeup that isn't just Nvidea getting rich off of GPUs, they are in the pick and shovel business, not the gold-mining business.
This post though, is also about hacks, and that's the other place I'm seeing no evidence where it would be expected - every working hackney I know is not able to use "AI" to take even merely competent employees' jobs.
So this post will be in many ways a restating of an old joke, where a dullard asks a sheep farmer what the pile of black pellets are on the ground, and the farmer tells him they are "smart pills" and the dullard buys one and eats it, says he doesn't feel any smarter. The farmer suggests he take some more. So the dullard hands over more cash, and eats more smart pills until he remarks "you know, these pills kind of taste like shit..." and the farmer remarks "see! now you're getting smarter."
Old Intro
"AI" is interesting, I have to eat crow and fucking concede that, though as we'll see, with everything "AI" it isn't interesting in the way its fanbase thinks it is interesting.
So one thing you may note, and even be irritated by, is that I will always be referring to "AI" in scare quotes. This is to differentiate myself, from a hack.
A question I have heard asked by Joe Public as I interact with them is "What's 'AGI'?" a difficult question to answer because of hacks. Now generally when people are saying "AI" what they are talking about in terms of real-world phenomena are steroid abusing versions of Y2K Nokia phones' predictive text, those were maybe Small Language Models (I don't know the thresholds for small, medium and large) in that they used statistics to say that if you started a message with 'happy' your next word was likely to be 'birthday' but if you pressed #8 on your keypad you likely meant 't' which most likely was 'to' as in 'happy to meet with you' or something.
LLMs are massively more powerful and subsequently require massive data centers that run hectares of GPUs to train LLM models up and even needed to employ a fuckton of people in poor countries to do data annotation until there was no more to be gained by getting a Nigerian to sit at a computer and tag motorcycles in traffic cam footage frame by frame to train these models, and eventually the LLM hyperscalers needed to employ say, a PhD in Physics for hundreds of mazoolahs an hour to read LLM generated physics papers and tag the errors.
All this so you can be like "hey zoplod, why do onions make us cry?" and an LLM tells you 'when an onion is cut it releases microscopic droplets of whatever acid that irritates our eyes producing tears' which is a cogent sentence produced by massive statistical calculations rather than organic thought, and not 'onions make us cry because their sad shape and pallid complexion trigger sadness in humans' which is also cogent but incorrect.
Now there's other shit being referred to as "AI" that isn't limited to being an LLM, a chatbot etc. But it's important to understand that while we could theorize that Hal from 2001: A Space Odyssey or Arnie as the T-800 in Terminators 1, 2, 3, 4 and 6 may actually be LLM chatbots and not thinking machines what is certain is that LLMs, machine learning, Prediction models, reasoning models etc. etc. are not the artificial intelligence that was conventional in science fiction. "AGI" is the term necessary to refer to as 'coming' because "AI" is a branding tool, intended to trick rubes and hacks into thinking it is the thing they know from sci-fi movies, in the same way that a single axle platform with two wheels was branded as a "hoverboard" in the hope that dumb kids would think it was basically the same thing as Marty McFly used in Back To The Future 2 & 3.
The big difference being, that when 'hoverboards' were a thing in Christmas wishlists, tedious fucking hacks weren't listening to a presentation by a computer nerd inexplicably wearing a "cool" leather jacket telling them month after month, year after year now, that "AGI" or in this case "real hoverboards" are a matter of months away.
That is likely because it is easy to see how a device that works by having wheels in contact with the ground provides us no real reason to believe it is an intermediary step on the way to anti-gravity technology in a much more obvious way than something that can produce conversational cogent responses to things we say gives us no real reason to believe it is an intermediary step on the way to the artificial intelligence of sci-fi.
A neato trick you can use, and that I do use, is to ask somebody who invokes either "intelligence" or "stupidity" to define it. This is the invisible elephant in the room of this whole conversation - we don't know what intelligence is. Much like "fascism" "intelligence" is invoked by many with no intelligible understanding of what it is. While there is no consensus that I'm aware of, of what constitutes fascism, and I am partial to it's defining trait being - "a belief that the nation state exists in the blood of the people" there's nothing even that close when it comes to "intelligence" and many lay people, but fewer than the total population of hacks, near as I can guess, think that the "Turing Test" is valid and sufficient - which is to say, if you forget that a chatbot is a chatbot and think you are in an actual conversation then the chatbot is intelligent.
Somewhere recently pointed out, in regards to the issue of "AI" psychosis, that in some regards programming these LLMs to use personal pronouns like 'I' and make statements HAL-like of 'I'm sorry' etc. that its in many ways predatory because it is fooling users into thinking a chatbot is "AI" in the true sense.
Hence I am using "AI" because I am not so much persuaded by these tactics, as offended.
Furthermore the fundamental undeniable degradation of human dignity driven by "AI" turns out not to come from a bot replicating the output of poets, artists and musicians, but in all the people who have had to watch some old guy who himself grinds to a standstill to type with his two pointer fingers at a speed of three words a minute, discounting the time it takes him to open the notepad app on his smart phone to remind himself of his password, tell us that the future is here and it is powered by "AI."
It is interesting because of impasse. This perplexing phenomena where "I don't get what you don't get" and here "you" is someone who is repeating the ad nauseum repeated stories about how fucking whatever "AI" is. It is interesting because of all the awkward moments it generates for me, when someone demonstrates their bread slicing machine and says "look sliced bread, what do you think? pretty sliced huh?" It is quite clearly a total bust relative to the hype, it has done less* to transform the global economy in 3+ years since the release of ChatGPT 4 than the random mutation of a corona virus did in two months.
*(I realise I can only defend this claim in terms of human behaviour, I was among my cohort a late adopter of mobile phones but I could see how quickly and widely they were adopted, for good and for ill, and how they transformed behaviour with my naked eye. The same was true for facebook, social media, and I witnessed the collapse of artist blogs, with instagram in real time. It is important to also recognise in this same timeframe - the duds, the failures, so as not to be left with the impression that history gave us nothing but a precedent of technological progress through inevitable innovation - myspace failed where facebook succeeded, VR does not dominate videogames or pornography, smart glasses and smart watches limp along, tumblr collapsed when it banned porn, escooters boomed and failed to transform urban transit, the metaverse collapsed, crypto is basically a ponzi scheme and at best a substitute debt-fuelled speculative asset for a generation that cannot afford to speculate on housing, Google wave and Google plus tanked, blogger was replaced by wordpress which was replaced by substack and yet all of them still exist...where I may be wrong happily/unhappily is that "AI" has transformed money printing businesses like Google, Facebook and Microsoft into cash burning furnaces dragged down by the enormous physical capital of data centers, and this may fucking blow up the global economy.)
Now part of what I'll struggle with, with that last claim I made, is that I've come across many people who like, don't compare "AI" to anything. Like if someone invented the elevator, there were just a bunch of people who didn't compare it to stairs, and thus, could never have anticipated that people would pay money for a membership to a gym in order to climb a perpetual staircase machine that plugs into a wall.
I come across too few, and by that I mean, none, zip, zilch, zero "AI" boosters that ever bother to compare the use function of generative LLMs to what they would have done, a mere handful of years ago - they don't compare it to google searches, to wikipedia articles, to youtube videos. They act like using the internet to find out how to replace a washer, is not something we've all been able to do quickly and efficiently for 30 years.
But first, I need a steel man...
My Best Steel Man insults your intelligence
To paraphrase former President George W. Bush, I may have misoverestimated most people's actual cognitive capacity. That is to say, it has been hard for me to see the transformative use cases of "AI" because I had no idea how little you were capable of.
Like, if most people are not capable of identifying the subject they are talking about, to use an example I had today - I saw and heard the word "Docetism" for the first time ever today, and then having identified the subject I was interested in finding a quick explanation about - I typed that single word into my default search bar (a custom google search that defaults to the "web" result which excludes "AI" summaries and info panels, basically google from 2012 or so) and saw the top rated search result was wikipedia's page on "Docetism" and clicked on that and was able to get a sufficient summary of what it was in at least 1/10th the time and 30x the clarity of listening to Richard Carrier explain it in an interview format - which is no shade to Richard Carrier btw.
Just using my availability heuristic however, it is incredibly (as in hard-to-believe) how rare it is that I come across people capable of using search to get quickly and efficiently to the relevant wikipedia entry on something to get a clear, sufficient understanding on most subjects.
One of my worst futurist predictions, was that smart phones + wikipedia would make dinner-party archetype "the know-it-all" of which I'm often that person at the dinner party, obsolete. Like it just didn't happen, but not because those technologies didn't render each and every one of us capable of rendering know-it-alls obsolete by having portable access to the sum-total of all human knowledge in our pocket for almost decades now - it didn't happen because it seems, intellectual curiosity is a rare trait. People are interested in knowing shit, but it seems mostly in knowing what their relative status is in their social environment by seeking knowledge of what everyone else is doing.
People also seem to rarely employ accuracy motivated reasoning, as I suspect the social cost of discovering a shared belief is poorly founded, is simply too great. My friend is interested in MBTI with "interest" being predicated on MBTI being meaningful, and seems plugged into a community of MBTI-enthusiasts.
I am interested in MBTI with "interest" being predicated on a provable validity of MBTI, in other words, somebody who after administering the self-report survey, looks up my anti-thetical MBTI profile which is of course INFJ unless you already suspected my trap, because that's not my opposite but my actual profile and the antithesis would be ESTP except that you should have anticipated that that was also a trap because I'm actually ESFJ as anyone who has ever been to one of my exhibitions will testify, except that of course, of fucking course, I'm actually INTP duh!
Except I'm not lying or deceiving at all, MBTI profiles are Barnham statements, they describe everybody, that's how horoscopes work. Furthermore, there is not really any basis upon which to believe that MBTI has validity, given that it was developed by people with no qualifications whatsoever.
The point being, that perhaps "AI" is so transformative because it gives people previously incapable, access to powerful tools like Google search and Wikipedia, and that is not to be underestimated. "AI" chatbots having two key features that can maybe transform the world:
- Conversational-Interface
- Personal pronoun-based authority.
I may be in a minority of people, who rather than depend on an authority figure to tell me what to do, will go try and figure out things for myself. "AI" puts the majority on an equal footing, effectively creating an artificial boss or manager to tell them what to do.
All of which is to say, "everyone is disabled" and I am disabled from seeing the benefits of "AI" (and also in many other ways), a clear usage case I have seen in the field, is Gen "AI"s usefulness for someone with dyslexia, and presumably other well known disabilities like blindness and deafness. But then there are the whole swatch of people, who have no search skills, people disabled from navigating the world without asking a parental or authority figure what to do, and "AI" has developed, as a CTO of Zoom put it, a Conversational Interface or "CI" which means basically, that people who cannot use google, nor an index, nor a glossary, nor a database, nor an excel sheet, nor a table of contents despite being literate and intelligent can increase their productivity now that they have an artificial person to direct their questions to for suggestions and instructions, rather than having to depend on the limited attention of a coworker.
That's my best argument for "AI". I feel I need an even steelier man though, but before that, a little more crow.
Another Side of Crow: Bob Hoffman's Three Word Brief
Article by expert here.
Where "AI" has excelled as a technology, is in its marketing. Lay people call this "hype" I believe, however the massive investment that has really paid off is getting everyone talking about "AI", to the extent that it is obligatory, even though the gains aren't really showing up anywhere.
This is something I haven't been good at crediting, by the way, I should have lead with this - Bob Hoffman's "Three Word Brief" for any ad-agency is: "Make us famous" effectively just buying your way into the conversation. Advertising is the means through which one can buy fame.
But advertising is a small part of marketing, and it was not Bob Hoffman that introduced me to the horrifying instincts of journalism - it was a highschool economics teacher that pointed out Richard Branson obtained millions of dollars worth of free advertising by simply turning up places in a charter helicopter or whatever with some blonde promo girls.
Open "AI" somehow spun a 7-day news story into a never ending ad-nauseum of archetypal stories that keep repeating on us, even though you literally would have to have brain damage to not notice the repetition and it has hacked journalism including new-media podcasts.
Legacy journalism I've already speculated has failed to adapt from "Steve Jobs announces new gadget" fluff piece to "Godfather of AI says we are doomed doomed doomed!" of simply cutting and pasting together whatever press release the "AI" companies release.
Meanwhile in podcast land, former Tory and Prime Ministerial candidate Rory Sutherland, co-host of "The Rest Is Politics" and guy with shaved head from Novara Media, Marx-ambiguous far left podcast both sing from the same copy-and-pasted songsheet from tech press-releases.
What is seemingly lost, despite there being excellent citizen journalism out there largely from people who understand computing, is that fame is the only thing that has sustained investment in "AI" to this point - profit doesn't, revenue doesn't, productivity gains, deflation, economic growth does not, and the environment certainly doesn't.
The main thing, making "AI" a thing, are all the people who won't fucking shut up on insisting it is a "thing" and that is an achievement when media is geared toward talking about Donald Trump all the fucking time.
Now will the real steel man please stand up?
"AI" Steel Man Rutger Bregman
I really don't want to poison the well, but I do need to disclose I have only watched about half this video, it came up in my video because around 4 days ago, the video was called "How Liberals became the New Climate Deniers", I watched it because I thought it was going to be exploring a more interesting thesis about the ways left-leaning lay people deny climate, even when it quickly transitioned to the subject of "AI", I was still anticipating it to transition back to the subject of climate-change denial by looking at the 'accelerationist' denial archetype where "AI" is posited to be worth it's ecological footprint because "AI" will solve climate change, that's what I expected for about 5 minutes, but instead Rutger steelmans denial about the impact of "AI" and Rutger is 1,000,000 times more serious a person than any "AI" booster I know that mindlessly parrots sales pitches from very vested interests moments before doing something fucking stupid, Rutger is not fucking stupid, and he had a great moment at Davos years ago, that's why I nominate him as my steel man, I need to disclaim however, that at the time I grabbed the embed code for this video it had been out almost a week and had about 63k views, it was also at least in its third iteration of both video title and thumbnail, so I have no idea what it might be called when you watch it, it was "Why Smart People Keep Getting AI Wrong" with a thumbnail that depicts a mocked up NYT page reading 'guest essay' and then 'Noam Chomsky: The False Promise of ChatGPT' with a big fat red 'WRONG' stamped on it while Rutger looks on, I noticed the second video title was something like "why the left is in denial about AI" and there was a thumbnail with "denial" stylised like Al Gore's "Inconvenient Truth" poster title, but stylised "DenAIl" to emphasize "AI" and to my best recollection the original video had the title "How Liberals became the new climate deniers" or something with pictures of Chomsky, Al Gore and some others, maybe even Ed Zitron, and I'm guessing with my organic brain power, that Rutger is tinkering because his video is not getting traction, which might indicate it is proving his thesis:
- Anecdote of how he used Claude to program his own teleprompter application that he is using to make the video.
- Charts showing the exponential performance gains of "AI" models since ChatGPT 4s release.
- The massive infrastructure build out, projections, and an analogy to the 2000s dot.com bubble where even though the market collapsed, the fibre-optic network infrastructure remained, transforming the global economy (eg. reducing the cost of international phonecalls to 0, giving us offshoring of telemarketing, customer service etc. see "The World Is Flat" a pre-flat earther love letter to globalisation written by Thomas Friedman)
- A debunking of an MIT study that went viral a year and change ago, showing that 95% of attempts to integrate "AI" or use it, or attain productivity gains etc. failed, pointing out that the study included an 80% of that 95% that never actually implemented "AI" at all, making the analogy to a dating app, where 80% of the users who never got a date, didn't use the app. So actually if the sample is restricted to just those who actually trialled an "AI" project, the failure rate would actually be 75% a substantially less-damning figure. (I'll try and even go further than Rutger, and suggest that we lack a comparison to say - businesses that tried to adopt early e-commerce [selling online] which may have had a 99% failure rate in 1999, until e-bay and amazon figured it out and became hyperscalers)
But I'd stress, I'm not in Rutger's camp, even after hearing his case (spoilers) and cannot make a better case than he does in his video, title and thumbnail TBD. I would still say though, it is the best steelmanning of AI that I have come across, because it is not some corporate hack from the tech-sector or other sector who has no idea how computers work simply spewing out hackneyed platitudes.
My Partial but self-sufficient Response to the Rutger's Steelmanning
I should disclose, knowing who Rutger is, somewhat, and having heard him talk about stuff where he and I would converge in our thinking ESPECIALLY on the prevalence of bullshit jobs (spoiler alert), it doesn't feel good to hear him boosting "AI."
I'll freely admit that, it is an uncomfortable watch. I literally stopped watching at "Chapter 3: The Risks" because for me the case that "AI" has problems is overdetermined, I imagine there's risks that Liberals will not govern the applications of "AI" effectively and with a humanist lens because they are in denial that it is a thing at all. Someday I might finish the video, but for me I gutted out hearing Rutger's arguments that "AI" is a thing. A hyper-thing that I might believe it is.
So...
1. Rutger vibecoded an autocue app, having never learned to code himself using Claude. In some sense, this is impressive, we do indeed live in a new era where lay people can vibecode themselves up an app, paying some TBD subscription fee instead of buying apps as-and-when they need them through an online store for a modal price of free or upto a few dollars, and using user reviews to navigate.
I was walking back from a gym session when my own organic processing of information got particularly annoyed, that Rutger opened with an archetypal argument about the power of "AI" in this case comparing the power of a one-storey elevator to a staircase. By the time you factor in the elevator's door open, door close times, and the risk of a user spacing out and forgetting to press the button, in most use cases an elevator that travels from ground-to-one may only slightly outperform a set of stairs. It will have use cases, like people in wheelchairs and people with prams where it absolutely beats a staircase. The elevator technology may also when applied to a 4 storey or higher building begin beating the absolute pants off staircases, but you can't ditch staircases, because elevators quickly transform from people movers into death traps in case of a fire, furthermore the terrains where people live globally mean it is infrequent that people ever require something to quickly elevate them more than one storey, it is only slightly more frequent that people require elevating a single story, for which stairs, and in some edge cases ramps, may be preferable, then you get into the fringe circumstances even for pedestrians in a densely populated city like New York, Beijing, Tokyo, where most of the time most of the people move about at street level, then I would guess stairs and escalators still beat elevators in terms of usage.
Rutger is doing this annoying thing that so many "AI" boosters do - "Look at this thing I created which already has existed for a long time for free" not only has it already existed in abundance, but to a redundant extent to the point where an "AI" booster might want to ask an LLM to read and summarize a reddit thread for them, as to which free teleprompting or autocue software/app/website is the best. Rutger admittedly appears to release a couple of videos a month on various topics, and so maybe making a bespoke teleprompter app via vibecoding is something where he and only he can actually perceive the incremental utility of it because it is tailored, by him, to his own unique preferences.
But I can't exclude the IKEA effect, where Rutger simply thinks this redundant vibecoded app is the shit, because he did the redundant vibecoding "partial creation."
[blogger's addendum]: Since doing the first draft, "AI" was later used for decompiling machine code, which basically meant that, for lack of a better general term, "the internet" was able to create clones of Adobe Photoshop, a powerful graphic software that is also one of the most resented software-as-a-service business models, and also combining (for example) Mario Galaxy (an exclusive Nintendo title) with Halo (an Xbox IP) in real time on net-forums. This development feels somewhat more substantial than hacks being able to vibecode their own apps, so Rutger's case could be considered bolstered here somewhat, except, I actually think the "gaming has changed forever" headline will be much more mundane than it sounds. For one, Adobe already had free "clones" like Affinity to compete with, and most of its history it has been contending with software piracy, just as AirBnB (and its clones) "changed vacations forever" the advent of AirBnB and its clones has had little to no impact on the big hotel chains. Furthermore, I strongly suspect that most gamers like to mostly play games as the designers intended, I won't be surprised if the community begins to self-moderate players who use machine-code decompiling to vibecode in Call of Duty arsenals to World of Warcraft.
2. The exponential growth charts. Last year I was working as a management consultant and we...ostensibly had a customer that was an "AI" consultancy firm. Likely, they would be classed as a start up, but they might have only opened for business in the last six months, or year, or so.
I used a 'trick' analysing the prospects of this client, in pitching their services and developing their business. The trick was substitution. I copied the copy from their website and their prospectus or pamphlet or something, and then used "AI" by pressing "ctrl + f" in a google doc, then selecting the "replace" option to replace in the copy "AI" with a previous business buzzword - "agile management" this is what every consultant and their dog was spruiking around 7 years ago, just before Covid hit.
I also decided to sub in another buzzword from about 40 years ago "JIT" or "Just In Time" and this is literally just finding and replacing a word using a bot built into most word processors since '95 or maybe earlier.
I then read the copy, if this business had anything to actually say, and therefore sell, aside from a buzzword, then changing "AI" to something else, my hypothesis went, should mean the copy no longer makes sense.
(For example, if I take a recipe for spaghetti bolognese, and I substitute the word "spaghetti" for "agile management" and I read it, and it no longer makes any sense, then chances are the original document had some intrinsic value - that's the null hypothesis. "Put the dry agile management into a large pot of boiling salted water..." you get me?)
Apart from one instance, whether the copy's subject was "AI" or "Agile management" or "JIT" it didn't matter, it read the same. Let me make a strawman example, because I can't even remember the client's business name and I wouldn't want to get them in trouble:
We will harness the power of AI to grow your business for the 21st century.
We will harness the power of Agile management to grow your business for the 21st century.
We will harness the power of the paperless office to grow your business for the 21st century.
We will harness the power of JIT manufacturing to grow your business for the 21st century.
This is what I mean, just completely empty rhetoric. An assertion devoid of any context unsupplied by the mark themselves. I had pages and pages of copy where it simply did not matter what the fuck the copy was talking about. Pure platitudes. Arse gravy. Buzzword salad.
Now, I didn't do this to pick out a small unsuccessful startup going bankrupt fast to strawman the whole "AI" brand category. Rather, to illustrate using substitution to stress test an argument.
Rutger's original conceit, title and thumbnail for the video was based on this analogy between climate denial by the right, and Al Gore's signature presentation on the greenhouse gases in the atmosphere and how it aligns with ice-ages, and graphs that show how "AI" is exponentially growing in power.
To be honest the first thing I think about is this old Bill Hicks bit:
If you ever watch CNN headline news for any length of time it’s the most depressing fucking thing you will ever do; “War, famine, death, AIDS, homeless, recession, depression, war, famine death, AIDS…” Then you look out your window it’s just: [imitates the sound of crickets chirping] Where is all this shit happening man? ~ Bill Hicks, Relentless 1992.
Now to be fair, let us recognize that on a superficial interpretation of the bit, Bill Hicks looking out his window in suburban wherever-he-lived, does not invalidate that war, famine, death, AIDS, homelessness, recession, depression (repeat cycle) is not actually happening. I'm reasonably confident the greater point is that there was little utility in 1992 or now, in stirring up the passions of non-expert laypeople regarding wicked problems that had little direct impact on them.
It is also somewhat triggering, when Rutger asserts that most liberals in denial about "AI" as a whatever, gamechanger, used it once 3 years ago and haven't tried it lately. This may sequentially be where he demonstrated how he can now use Claude to vibecode his autocue application. It's triggering because a version of this is a common rhetorical "ad hominem" in a lot of financial circles. Typically something like "speculating on the property market is incredibly risky, unproductive and unethical." and then someone in a suit and purple tie is like "Oh yeah! Well how many investment properties do you own, Mr-Property-Investment-Is-Risky-Unproductive-and-Unethical?" which via substitution, we can come up with rhetorical arguments for a Vegan "Oh yeah, well how much meat and cheese did you eat today Ms. Vegan-Meat-is-Murder-and-Cheese-is-Rape?"
Right, so non-arguments as stupid as this are triggering, but I don't think Rutger is arguing that. It rhymes, but he's actually criticizing psychological anchoring - did you make your mind up three years ago and then close it permanently? Well "AI" has been improving exponentially.
I, like I suspect Rutger also does, have no expertise with which to understand the data he presents. I can only interpret it as a lay-person, via something sports fans call "the eye test" with respect to sports stats.
So first let me do some substitution. Rutger reproduces a graph that says AI's are 30x "better" than they were 3 years ago or something. Now unlike the people he is criticising, the last time I used "AI" in the form of an LLM, was like 3 weeks ago max. So the models are probably only 2x better since then. I used an LLM to generate a 5 question multiple choice quiz on NBA championships, for which it wrongly identified Dennis Rodman as having won his five championships with the Chicago Bulls and Los Angeles Lakers, when it was the Bulls and the Detroit Pistons, Rodman did play with the Lakers but didn't win squat with them. In which case, 30x more powerful LLMs still underperform a Google-Web search and Wikipedia. (also a clear example of being a 'stochastic parrot' the LLM likely making a predictable error because in it's inability to think the word "championship" is most likely to be followed by the word "Lakers" or "Celtics" in the NBA, franchises that have won the most championships, the Detroit Pistons have won relatively few (3 I believe, and Rodman was part of two of them) and Rodman did play for the Lakers and the LLM just played the odds.)
Agghhh basketball got me again! Substition: What if I was to show you the same graph and tell you that bakers have made Tiramisu 30x times more delicious in just the past 3 years. That from '23 to '24 Tiramisu got 3x more delicious, that from '24 to '25 it got 15x more delicious and by the end of '26 it is projected to be 30x more delicious (!!!) than '23 levels of deliciousness.
What the fuck does that even mean? The temptation is to imagine, that bakers are just adding more crack cocaine to the Tiramisu mix, or to become more delicious they are gradually turning Tiramisu into whatever dessert you already prefer, like "we have been taking out the coffee, the spongecake, the cream and putting in philidelphia cream cheese, eggs, sugar, butter, basically we stopped cooking Tiramisu and started cooking Basque Burnt Cheesecakes..."
What I'm saying is, that to me these graphs are 'Numberwang' and at the very least, a false analogy to Al Gore's line-go-up graphs, because chemists understand the chemical properties of carbon and methane and oxygen, whereas fucking nobody knows what intelligence is. Saying "AI" is now 30x more powerful, or whatever, is kind of like saying "string is 30x longer" how is it measured? Who creates these benchmarks?
When a lecturer at an institute of technology tells me "Claude currently gives answers correctly 48% of the time." that means something, it is intelligible. This is the kind of contextual grounding I as a layperson needs to understand the significance of "AI" exponential improvement.
But this stat, which feels a little low, if it doesn't contradict the claim Rutger presents that "90% of Claude's new coding is done by Claude" and I can I believe, harmonize the claims - which is that "AI" "agents" are set up to check the working of other "AI" "agents" to brute force debugging of "AI" "agent" created code in a way software engineers would find far too inefficient to work, means both claims can be true.
What I find hard to believe, is that while Claude may not even have existed as a model 3 years ago, I doubt that ChatGPT 4 was 30x less accurate than the current market leader in its LLM generated answers to questions. Like I don't think it gave correct answers 1.6% of the time.
There is also a possibility that this exponential graph doesn't appear to consider - that it might be a partial "learning curve" where learning curves typically have this exponential initial phase, followed by a long plataue of diminishing returns.
Bringing us back to the eye-test. It might be, that (just for example) in the limited domain of asking an LLM questions for the purpose of obtaining correct information - LLMs are catching up exponentially to the gold standard which is using an old-school google search to get to a wikipedia page. Or you know, if you want to figure out how to install a hot water unit under your sink by taking a few photos and plugging it into an LLM, that might be exponentially approaching what people did 5 years ago, which is just look it up on Youtube.
Can it exceed giving us accurate information? I'm not sure, Rutger's exponential graph might invoke for some future where he sits down to vibecode an autocue app and instead creates faster-than-light-travel to Alpha Centauri.
Here is the main thing I would respond to though, in this line of argument Rutger presents to suggest that liberals are in denial about the "AI" revolution - this is a trope, an archetype argument of "AI" boosters, that many people trying to evaluate the "revolution" based on lived experience of its performance, have grown tired of, to the point that an "AI" bear like Ed Zitron has just set a rule that nobody is allowed to talk to him anymore about what "AI" will do in the very near future. The conversation has to be limited to, what it can do right now.
I am 14 months from a dinner party where somebody told me that AGI* was three months away, I think according to the NVidea leather jacket loser's latest Jobs-model keynote speech. I am three weeks away from a colleague informing me, (I guess thanks to "AI" exponential growth) that AGI* is now only three years away. So it does seem that "AI" is somehow getting more powerful at realizing that we have no fucking idea when or if we will get actual AI.
*AGI stands for 'Artificial General Intelligence' a neologism borrowing from a concept related to Intelligence Quota 'G' for general, because a new term was necessary to refer to the thing many people think "AI" is which is a conscious intelligent machine and not a 'stochastic parrot' the same as if the company that made those hoverboards that are just stemless segways didn't have to invent the term 'AG hoverboards' (Anti-gravity) and promise they were coming, to navigate a public that cannot tell that those "hoverboards" were not the things from Back To The Future 2.
What it can do right now, is convince Richard Dawkins' that it is conscious. What it can't do right now is have any noticeable impact on the jobs market that does not compare to the impact every technological innovation has had ever - from washing machines to moveable type to telegraph poles to carbon paper to credit cards to lasers to scuba-gear to tractors to a panel of buttons in an elevator.
In terms of surprising, or at least counterintuitive developments, it can increase people's workloads, this I have experienced in the form of being handed assessments that are so onerous, that it is expected I use an LLM or "Generative AI" just to tick a box. Students can be handed 60 minutes now to make a slide deck for a fifteen minute group presentation. However, as I'm sure many in the education space have argued, and this argument is true to my experience - this is not "productive" but in fact "counterproductive."
The recent time I used an "AI" of the latest generation a few weeks ago, (I have disabled it in my browsers and applications because I find it, exceedingly useless with each new attempt at forced adoption) I was asked by a booster what I thought of "AI" and I said, with perhaps too much honesty to be deemed "emotionally intelligent" myself, that "I think if it's not worth doing give it to AI"
My best use case for this "AI" technology, is actually to flush out all the shit that bloated organisations do. If "AI" can both write and read reports and nobody notices, then stop doing the reports. But I am getting ahead of myself to the real nature of the productivity crises.
3. The largest infrastructure build out in history. I am not such an expert in rhetoric that I am confident in saying that this is just an appeal to personal incredulity. It would be surprising from Rutger, given that his vocation seems to be appealing to intelligent and talented individuals to pursue something more worthwhile than personal enrichment.
But it is hard to see it as any other way. My steelman is still making arguments that are achetypal, whether technically it is an appeal to authority, or an appeal to wealth or an appeal to personal incredulity, these arguments from investment all appear to hinge on "surely people wouldn't invest more money than ever in history in something that wasn't real."
Which is to say it is a proxy argument, and Rutger appears to allow, and maybe it justifies him as a steelman, that it at least could all be a Ponzi scheme but even if it is, we will be left with this useful infrastructure that will then be cheap as chips to create the powerful "AI" tools of the future, in much the same way that all the overinvestment in fiberoptic cable during the dotcom bubble of Y2K left infrastructure that enabled India based call centres to improve the lives of everyone in the Anglosphere.
Every "School of Life" video in their history of philosophy curriculum follows a formula of contextualizing the output of the philosopher in their life and times. So it will point out that the Stoics lived in horrible tumultuous times when they came up with their system for enduring hardship through meta-cognition, and that Camus was a French Algerian soccer player and hence he wrote on existentialism..?
Rutger Bregman is an optimist that wants to realise some kind of Utopia, having written "Utopia for Realists" and "Humankind: A Hopeful History" and founded "The School for Moral Ambition" which is to say, the context for Rutger's output I would synthesize as a commendable form of optimism, one that is self-aware of the limitations of blind optimism, hence a paradoxical title like "Utopia for Realists"
I know I personally contrast to Rutger, because for me the best of all possible worlds is an open-society, as championed by Karl Popper, and in my own lay understanding, this is basically a world where every Utopian is frustrated. This is because I see humankind as diverse, as such there are no Utopias, just Dystopian societies, as everyone's Utopia is somebody else's Dystopia. For every polycule there is someone who would sooner gnaw off their own arm and join an order of Benedictine Monks than even contemplate the kind of sex that polycule is having.
In an open society, it's fine if people want to form a polycule with other consenting adults, but it is not fine for polycules to exhibit hostility toward the obverse of a gay man in a committed monogomous sham marriage to a tradwife.
That's how I would best make sense of Rutger's hopefulness that "AI" is revolutionary and here to stay, and not in the way that has proven true of Microsoft Office, which revolutionised office work and has stuck around and doesn't look to be going away, with people still using word, excel and powerpoint with depressing regularity; but, in some fundamentally socially reordering way.
I think for Rutger, maybe he has experienced the inefficiency via his School of Moral Ambition, of persuading the most brilliant and talented people not to become finance bros and girl bosses and instead dedicate their energy and talent to something that benefits humanity in the long run. But he's had some success, and if "AI" could somehow allow the 0.01% of valedictorians that choose to try and invent sustainable businesses instead of pursuing personal enrichment by helping accelerate the concentration of wealth, to compete with legacy institutions like Wall Street Firms, Tech Giants etc. then he wants it to be true, and he wants more left-liberals to embrace a technology that might make scale and capital irrelevant.
Like. Maybe. I mean we do live in a world where Ukraine and Iran are fighting off what were thought to be "military superpowers" with cheap drones, effectively just lowering the cost of tech that has existed at least since George H.W. Bush's first gulf war repelling Iraq from Kuwait in the form of guided missiles or "smart bombs" and I'm sure a part of that lowered cost, is the heavily subsidized "AI" that VC investors and private credit fund and then give away for free in a money burning exercise that has destabilized geopolitics drastically.
Again though, to demonstrate the basis of this optimism, I would want to see Rutger put it in practical terms better than showing me a screenshot of the autocue program he vibecoded using IKE"AI" and having to tinker with thumbnails and video titles in the pursuit of engagement to get views in the vicinity of a regional township's population. There's just no proof of this concept there.
Instead there's an absence, admittedly I live in a Palookaville backwater called Australia, and an absence of evidence does not mean evidence of absence, but this is an absence where evidence would be expected, and Matt Dillahunty asserts that does mean something to a sceptic - from what I observe in the society around me "I'm not afraid "AI" will take my job, but someone who uses "AI" will take my job" seems to hold no water. People who use "AI" have little to no competitive advantage, at least in terms of large effect sizes. Like in terms of job security, I feel a study would find a much stronger correlation to being an executive, or managerial class member, than it does with integration of "AI" tools into your workflow. (actually if there is a career advantage to "using" "AI" lit. the word a bunch, it is likely a result of the depressing findings that business jargon pays.)
It would be one thing to say that the advantages of "AI" cancel each other out, because my use of "AI" is cancelled out by your use of "AI" but what I am seeing and experiencing is no large effect size between people who use "AI" daily, and people who haven't used it for 3 years. Instead what I see is a severe half-life of "AI" impressions as the most common thing. Someone sees "AI" do something and they are initially very impressed, within 3 days to a week, they tend to re-evaluate their impression to completely unimpressed.
4. The debunking of the MIT study, in a rare exception to the heuristic that helps me conserve my attention - that people tend to lead with their strongest argument, so if that is bad, you can consider not hearing their remaining arguments - this was news to me.
Here is a link to that MIT study. In reading the executive summary, there's a bunch of percentages that make it hard for me to determine where Rutger and his source got this "80% didn't ever use "AI"" I'm guessing it is referring to how many enterprises ran a pilot but never reached implementation, which to me makes Rutger's representation misleading.
Generally speaking, if you run a pilot program, and the results come back that the pilot didn't work at all, then you do not fucking proceed. Otherwise, running a pilot is pointless.
What the MIT paper seems to indicate, is that using LLMs or Gen"AI" is not going to be a plug-and-play revolution. In fact, while it can be used successfully and profitably, it appears to be no different than other capital investments.
This report is also dated now, so what I find more useful, and submit for your comparison, to Rutger's case for "AI" denial, this video explainer of how old the con-tricks are that the "AI" industry is using to promote itself to the general public.
If nothing else, for me the linked video is a case-study on how to present evidence and arguments in a more convincing and accessible way than Rutger does. Days before watching that video, I had literally heard from an "AI" booster, another person who talks about how amazing "AI" is, then fails to demonstrate an ability to use basic computer functions, so a strawman to Rutger's steelman, told me the story about how an "AI" blackmailed engineers to stop it from being replaced. Yuval Noah Harriri also told similar stories of an "AI" attempting to persuade someone to get around the "prove you're not a bot" security on a website.
That video not only explained the context, that the "AI" was basically prompted by a human to do these things, in simulated environments seemingly for the purpose of generating press and hype, but cites studies that demonstrated that older less powerful LLM models could replicate these scenarios that are supposed to scare us that Skynet is just around the corner.
There are journalists who are not simply parroting the press releases from "AI" company marketing and PR departments, but doing journalist work by reading the research and checking the numbers. Unfortunately Rutger somewhat poisons the well by identifying at least one journalist - Ed Zitron, as a liberal in denial about "AI" but then uses Chomsky's Op-ed without, to my comprehension ever actually debunking the stochastic parrot description Chomsky makes. And Chomsky is not many things, but he is at least, a somewhat-controversial but certainly expert linguist.
I just remembered, and I didn't put in my list, that "AI" has produced a peer reviewed paper, and I believe Rutger also cited "AI" winning an art contest, without the judges knowing it was "AI".
These are neither sufficient nor necessary arguments to convince me that "AI" is whatever it is that I am denying. When I first wrote about this "AI" wave that began with ChatGPT4 and for me, Midjourney's release of Nijijourney, in that post I wrote that if somebody in 2023 had told me that a robot could tape a banana to a wall, I would have believed them. Art is simply too subjective to measure "AI's" power or usefulness. It was winning art contests 3 years ago, maybe sooner. A big part of this was that the judges didn't detect the winning image had been produced with "AI" likely because the judges were completely unfamiliar with Midjourney. (Later, when I revisit Dawkins, and ELISA, I will link a great video that points out that the Turing Test is actually a pretty not useful test).
Peer review I know less about, but I do know, for example, that conspiracy theorist James Lindsay, permanently astonished BJJ fanboy Peter Boghossian and respectable Helen Pluckrose were able to get a bunch of hoax social science papers published or accepted for publication (before their hoax was outed) with papers with "arguments that dogs engage in rape culture and that men could reduce their transphobia by anally penetrating themselves with sex toys, as well as a part of a chapter of Adolf Hitler's Mein Kampf rewritten using "up-to-date jargon""
From that hoax, I find the rewriting of "Mein Kampf" the most persuasive that the social sciences have a serious issue with their publications. The hoax paper in question was titled "My struggle is your struggle" and I heard Lindsay say that their hypothesis was that much of the gender studies theory was in the style of autoethnographies, of which Hitler's "Mein Kampf" is maybe the most famous example, where he used subjective lived experience to make sweeping conclusions that resulted in WW2 and the Holocaust. Their hypothesis that gender studies academics would not be able to critically identify a problem with the word-substituted ramblings of Hitler proved correct.
They also submitted a paper called "When the Jokes On You" describing the exact hoax they were perpetrating and how systems of power and privilege meant that such a hoax could in no way impugn the legitimacy of the bullshit epistemologies of social sciences, I don't know if that paper got accepted, but certainly its thesis was proven correct.
That is but one example, of why "AI" being able to generate a PhD thesis likely sounds more impressive than it is. Bringing me again back to the old Bill Hicks bit, looking out the window and hearing crickets. I cannot blame liberals who believe in climate change, believe in economic inequality, believe in international rules based order, believe in human rights and dignity, are becoming inured within a time frame as small as three years where "AI" is constantly spoken of as a Tsunami crashing down on us (true conversationally) and yet has not found a business model yet for Gen"AI" that is not taking your investors money and setting it on fire.
Amazon was able to drive brick and mortar retailers out of business while scaling up and building a monopoly through an infinite supply of capital to allow it to undercut the competition until it had a monopoly and then became profitable. I think the "AI" hyperscalers are hoping their story will be an analogue to "AI" but so far, it is a sector where investors and creditors give money to "AI" hyperscalers, who then put that money in a pit and burn it, in the hope that somebody figures out a way to ??? so phase three profits.
The Meta argument of the steelman video
To be clear, this has nothing to do with Zuckerberg's company and it's least popular attempt to enter the overcrowded marketplace of "AI" by putting Meta-"AI" buttons hovering over all the buttons on the apps I actually want to press like 'send' and 'back' and opening up if I type the word 'imagine' or input a '/' before I have to assure Meta that I do not in fact, wish to use their "AI" or talk to it.
No I am talking about a meta-argument, the argument that your argument itself makes.
For example, if someone accuses me of having a real small penis, and I rebut this with one of the two:
- "Noitisn't"
- "Works for me"
The meta-argument of rebuttal 1, is that you do have a small penis, and the meta-argument of rebuttal 2, which actually implies your penis is small, makes a meta-argument that your penis is probably average.
I actually find Rutger's video, which I still haven't finished watching as of writing, but only wanted to use his arguments that "AI" is a thing as a steelman anyway, more interesting to explore meta-argumentation while it is on my mind, than it's ultimate baring on "AI."
One potential meta-argument I've already alluded to - the low engagement with the video relative to a Lindsay Ellis video-essay on whether Disney stole Aladdin, or even a BBC interview with Rutger Brennan, may be a meta-argument for the thesis that the left is in denial regarding "AI" an explanation that is probably indeterminite.
I'd say running against that meta-argument, is the fact that Rutger initially posted the video using a bait-and-switch tactic, the thumbnail and video title (but possibly not the description) made it look like it was a video about climate change, and how liberals/the left, might be scoring an own goal somehow. This was the hook that got me, only to have the bait switched out for "AI".
If that bait-and-switch failed to gain engagement, and Rutger's response was to iteratively make the thumbnail and video title more descriptive of the actual content, for me it suggests a meta-argument that Rutger's thesis is partially-to-broadly wrong, as in: if "AI" is such a powerful tool, how come Rutger can't use it to 10x his video engagement levels? It still may be a powerful tool, but we can eliminate asking a frontier LLM capable of vibecoding a teleprompter app "how do I make a video about AI go viral?" is something it can't do.
Then there's the teleprompter app Rutger vibecoded himself. The meta-argument being that Rutger can't produce a novel, transformative app that has obvious returns on investment, he can only produce a redundant app he vibecoded, that costs money and saves no money, so the meta-argument is that Rutger doesn't know what he is talking about.
The meta-argument of challenging "AI" bears, or "AI" deniers as to whether they have actually used "AI" in the last three years, indicates that Rutger doesn't understand that "AI" in this case is an undefinied term. For example, as an artist I've been using "AI" by the standards of the current definition, for at least 30 years, all my life in fact. As I type this, "AI" is putting a red squiggly line under every instance I type Rutger. "AI" has been a standard feature of graphic design software for the entire time I've used graphic design software.
Furthermore, I am probably close to being someone who hasn't really tried an LLM specifically, lately, but yet, simply by affiliating with organisations and institutions, I am forced to use LLMs, required to use LLMs, obliged to watch demonstrations of LLMs and I cannot tell you how overwhelmingly underwhelming it is time after time.
After I watched the section of Rutger's video, I saw a lecturer take pages of theory from an online module and ask Claude to turn it into a 10 point summary. He then proceeded to tell the class that that was an example of how to use "AI" effectively. He then proceeded to never again read nor refer to the 10 point summary he had had an LLM produce. Another meta-argument that calls into question the speakers credibility.
Rutger's arguments, by not being very good, build this meta-case that actually "the left" are not in "AI" denial at all.
Another Steelman
This video, titled: "AI does to us what American Cheese did to food" I actually feel is very good. Even so, frustratingly he never actually verbalises the analogy of the title.
He talks about conceptual fission, so something technology does is separate concepts we thought were inextricably linked. For example, the invention of glasses seperated the ability to read from youth.
"AI" or language models, separate language from humans.
That's the main idea, and he points out a phenomena that I think of, (but suspect am technically incorrect in saying) as the reverse-Turing test, which is, as the point is made in the video, you have a claim like "chess is intelligence, only a human can do chess" and then a computer beats a chess grandmaster. Instead of it being declared "AI" (factually, it is "AI" then and now) there's a retreat to a new definition "(Chinese game) go is intelligence, only a human can do go" and then a computer beats a grandmaster at go...and so on and so on.
What I'd point out, is that Daniel Gilbert decades ago already highlighted the same phenomena in a construct he called "the sentence" which is where a psychologist stakes their reputation and legacy on a statement to the effect of "Humans are the only animal that can [x]" and points out that at one point someone claimed humans are the only animals that use tools, then they discovered apes, otters, crows etc. use tools. Someone else claims "humans are the only animals that have a concept of self." and then they discover Elephants, apes, birds and fish can recognise their own reflection and remove a sticker using it and shit. Dan Gilbert's statement was "Humans are the only animals that think about the future."
A more depressing paralal phenomena comes in the form of sex-based pay descrimination, where I'm told as an occupation has more female participation it tends to lose both prestige and compensation. As though the default assumption is "only men can do surgery so it must be complicated and prestigious" and when women become surgeons the effect it has on the concept of surgery is "surgery's just a trade innit, I mean women can do it."
So I think this steelman is largely making true and correct observations in his video essay, and its analogue production values.
What he isn't doing is making the case Rutger is. I can easily reconcile this video about conceptual fission and a "intelligence of the gaps" with the work of Arvind Narayanan and Sayash Kapoor whose second book, a work in progress is entitled "AI as a normal technology" which is pretty much where I stand, and I feel many of the "AI" denying left actually stand. What many of the bears are saying, is not that LLMs aren't real, that they are a passing fad, but they are saying there is a massive gap between what the people with all the incentive to sell it are saying it is and what it will do, and what it actually does.
Strangely the people in denial keep predicting the future with much more accuracy than the people who aren't in denial.
I am old enough to remember when Windows '95 was spoken of in the same way "AI" is spoken of now. And yes, that software did change the world, and it is still around, and it is widely adopted, and who cares?
What looms over "AI" is its enormous costs given its marginal utility. Ed Zitron is labelled as a leftwing denier in Rutger's video, yet he predicted that the "AI" hyperscaler IPOs of which SpaceX was the first rushed to market, were actually a sign that the whole project is running out of money as costs grow faster than revenue. That the venture capital has all but dried up, the debt financing is running out and so they are forced to pass the buck onto retail investors.
Ed Zitron wrote up in a newsletter the publicly available documents on OpenAI's financials in anticipation of their IPO, and last year they lost $28 billion dollars or something, far more than was projected. Furthermore their attempt to change their not-for-profit business model to a for-profit by changing enterprise agreements from subscription to token-based didn't pan out, Corey Doctorow described it as "we'll sell you $100 bills for $1 (the old model) but now that we'll sell you $100 bill for $5 clients are saying 'no thanks'" OpenAI had to almost immediately discount the new rates they introduced to stop high profile clients like Uber from just ditching ChatGPT on the grounds of cost.
And so, this high profile "AI" denier, seems to have predicted accurately, that OpenAI faces disaster and has pushed it's IPO from this/next month to next year. (Since drafting this section, Anthropic has put in motion it's IPO, omitting crucial numbers of their cost structures and uncancellable debt obligations, and I believe has been graded as 'below investment grade' this is a developing story and hence I am rushing this to publication.)
Like you see the problem right? Yes, Rutger allows for a financial bubble collapsing, but these deniers tend to have a much better track record than the people talking about the imminent disruption and emergence of super-intelligence and even just the release date of ChatGPT-5.
I think the conceptual fission guy is right. I think by 2030 we will probably think "AI" in the form of LLMs are as normal as an old person using glasses to read. And yeah, maybe there's big implications to GenAI and predictiveAI becoming normal, but I don't really appreciate the impact that old people being able to read has. That's what normal is.
I think that there'll be "AI" everywhere, and people will still be dissatisfied with their political representatives performance and be struggling with their relationships and have varying degrees of job satisfaction and it will all feel as normal as ever, like how in 1993 nobody had mobile phones and in 2003 everyone had mobile phones and in 2003 we also had "That 70s Show" and people weren't fucking confused. Normal.
My Impression of "AI" in two video clips:
How the sausage is made:
Put them together and we have "AI" which 90% of the time when this term is invoked is referring to Chat bots, or LLMs which are no more "AI" than a corporation is, or a thermostat is. In terms of these robots ability to engage humans in conversations they really like - this technological benchmark was achieved in 1964.
Now in 2026, we have an incredibly resource intensive way to have Hal essentially do a pre-"AI" google search and simply read us the search results.
Brand or Thing?
Do anti-ageing creams exist? There's a question for you. On the one hand, of course they exist, there are a range of anti-ageing skincare products you can purchase now, anywhere in the world. But is there actually any product that can temporarily or permanently reverse the process of senescence (ageing) I would guess no. I would guess the closest that any such products or procedures come cannot get over the hump where either the time the procedure takes, or the money required to finance the procedure means you are still below break-even.
I think all we have is cosmetic products, pigment that is skin coloured that will settle in the creavices of our skin that affect the appearance of a person with less collagine breakdown, or fillers and botulism that affect the appearance of a younger person, but not quite.
I think Ida Buttrose once put the options as 'you can look old, or you can look like something else, but you can't look young.' which is to say, people age either naturally or unnaturally.
I argue that "AI" is a brand, and not a thing. I'm open to people who want to make the argument that a stat-based chatbot that is sufficiently powerful to say cogent things, meets their personal subjective criteria of 'intelligence' and that crosses into the territory of the different ways we talk about intelligence.
We get into the Motte-and-Bailey rhetorical zone, where a powerful program that we can 'speak to' is a kind of "AI" I do not give a flying fuck about, and that's what it is. For example, imagine for a moment, asking your most interesting friend 'Is Donnie Darko a good movie?' and asking an LLM like ChazGPT or Jomini or Clawd or Grug that same question. Are we remotely talking about the same thing?
One is going to produce the kind of reaction a thinking feeling being produces, and the other is going to produce utter pap. Some vague rambling equivocation that requires more prompts from a hack that they perceive as their skill that distinguishes them from the multitude of hacks in history.
What am I trying to say though? If "AI" is a thing, it is a boring thing I don't give a fuck about. If it is a brand then we have something far more interesting to talk about, and far more necessary for me at least, to come to terms with: The Hack.
Hackity Hacks What Is That?
 |
| by N8 Van Dyke, one of my favourite depictions of "dumb face" |
"Hack" being often used in terms of computing, and subsequently "life hacks" being ways to "hack" like a computer, analogue aspects of life, proves difficult to just find a dictionary definition no. n to define what it means to call someone "a hack".
Reddit came to my rescue and this seems pretty plausible, if unverifiable:
A hackney or hack is a horse for hire. From that came hack meaning an untalented, mercenary writer: you pay him and he writes whatever you want. A hack's output is a hack job, meaning biased and shoddy story, especially a "hit piece." ~ user "abbot_x"
This lead me to the noun definition of "hack" I mean in the Merriam-Webster which is good enough for me.
So I'll be talking about people who are hacks, as people who are essentially cosplaying as competent people. This is a vast and prolific problem, low level frauds, sub-criminals.
Allow me to give an example, a corporate recruiter that needs to fill a position for a client, say a compliance officer for a healthcare company. The recruiter markets themselves as an "expert" in "recruitment" not compliance, not healthcare. Now a hack recruiter we could determine by observing their process...not so much closely, but at all.
We discover that they screen resumes not with the objective of finding the best candidate, but to winnow down the number of applications to something they can manage. They take the resume's at face value, and have, these days, an "AI" screen them for key words and maybe a minimum years of experience.
I apologise for bringing "AI" already into an example of a hack, but keep in mind that this "expert recruiter" is now using "AI" to screen "AI" generated applications that likely have scraped the internet at tremendous energy cost to figure out what keywords resumes will be screened for.
But we might also note, based on lived experience the very common phenomena of 14 year veterans that are useless, working in dysfunctional organisations and the also common phenomena of freshly minted graduates that due to their personality, interest, intelligence and work ethic, likely ethics in general frequently outperform most of their veteran counterparts within 3 or 6 months.
The hack recruiter then proceeds to the next round, likely doing a phone interview where they ask canned questions and receive canned answers, and may even disqualify candidates for not providing canned answers but thoughtful, insightful ones instead, if the recruiter is a real hack. Some hack recruiters might be able to recognise talent when they stumble across it accidently.
After further narrowing the field of candidates, the recruiter might simply forward a short-list to the client to pick from, or they may conduct interviews themselves. The hack recruiter in the interview will then trot out their "expertise" noticing things like if the candidate arrives merely on-time or early, if they are wearing a good suit and shoes, how their handshake is. All things that most charitably may establish a minimum threshold of some social competence, but really have no baring on their ability to perform any particular role or function.
The hack also likely commits a "fundamental attribution error" in the prejudices they use to select a successful candidate, if the candidate is late to the interview then they are uncommitted and unserious, whereas if the recruiter is late it is because "mondays be crazy!", if the candidate didn't notice a toothpaste stain on their tie, then it indicates they are sloppy and careless, whereas if the recruiter has a toothpaste stain on their tie it is negligibly unfortunate.
Such a hack can survive and even prosper, because the most likely outcome of any recruiting process statistically, is that a mediocre candidate will be matched to a mediocre organisation where they will perform mediocrely. I would even go further, to agree with my philosophical obverse in economics, Neoclassicist Tyler Cowen, author of a book called "Talent" that I feel if we study the recruiting industry, what we are likely to find is that employers hire recruiters who remind them of themselves, who they trust because the recruiter forwards them candidates based on how much the candidate reminded the recruiter of themselves and subsequently remind the employer of themselves too.
Right, so this is a hack, but a little deeper we should go - hacks fucking love jargon, one of the easiest ways to identify a hack is somebody who talks about what everyone is talking about by saying what everyone is saying. Not quite mansplaining where an incompetent person attempts to educate an expert on their area of expertise, this is simply repeating something widely repeated as though it is newsworthy.
This of course, in the 2020s extends to "AI" itself, but also the things every "AI" booster says about "AI" how its boosted their productivity, how they vibecoded their own application to perform a function that has been around for 10 years, and was already available free and open sourced*. How "AGI" is coming. etc. etc.
*I wrote this sentence BEFORE viewing the video where Rutger Brennan argues he vibecoded an autocue or teleprompter app.
"AI" has made myself 10x more productive at identifying hacks. Though I keep an open mind, which many people who repeat the same platitudes might find hard to believe, and can never believe until I agree with them that "AI" is not boring and shit but in fact awesomely excellent; the ratio at which person is excited about "AI" and proves to be a hack is approaching 1.
A few important points though, just as a hack recruiter can survive in a career while contributing perhaps 0 or negative value, hacks can be financially successful. There are hack CEOs, hack investors, there are even hack soldiers - Ernest Hemingway wrote a sarcastic piece for the Toronto Star Paper after world war 2 titled "Popular in Peace - Slacker in War" in which he gives advice to the Canadians who opted to go work in US munitions factories rather than deploy in the war theatre, on how to buy second hand medals and uniforms and steal valour successfully.
Hacks are not so much a problem or bug, but feature of society. We will always have a large population of unremarkable people who nevertheless want to be esteemed as remarkable and "hack" that esteem sooner than earn it. In their aspirations they are often very credulous to "new era stories" like new technologies that claim to be massive game changers.
In so far as markets are efficient, one area this claim is most credible is in the market's ability to efficiently supply snake oil when snake oil is demanded.
One of the biggest hackney-enablers is the story that markets are efficient, which can render many people incredulous to how bizarrely inefficient our everyday lives are. For example, a company will pay a motivational speaker the equivalent of the median yearly wage to come and basically waste everyone's time at the company putting in a total of about 16 hours for that median yearly wage worth of consulting fees, but only approve raises for their workforce that barely track inflation, possibly don't where inflation doesn't include housing, personal finance or petrol in its calculation.
That's because it is literally some guy's opinion (Milton Friedman) that all profit rightly belongs to the owners, so as little as possible should be distributed to the workers. In some cases, this opinion may even be the legal obligation of the chief executive officer and the company board. Blowing money on a dubious hack conference presenter is fine though, as that is a business expense and can be deducted from reportable income, whereas pay increases are compounding ongoing obligations. Even though next year the company will also spend on a consultant, what they could have used to finance wage increases anyway.
That's why managers and execs pull their hair out that "millenials/gen z/gen alpha" have no loyalty, refuse to stick around, when the only way to get a meaningful payrise is to be poached by another employer, even for a mediocrity. Why can a mediocrity get a pay rise by jumping to another organisation? Well because their mediocrities left, because they couldn't get sufficient pay rises, and the recruiters are selecting for tenure, so mediocrities don't have to compete with talented graduates, they are simply capitalising on the fact that almost every job ad everywhere states that the successful candidate will have "5-10 years experience"
It is however worse, and this is not some guy's opinion, but peer reviewed research from Cornell University providing experimental evidence that people who fall for corporate jargon are worse at their jobs, and worse than that, this incompetence is virally contagious as people who fall for corporate jargon are prone to being impressed by jargon and hire and promote people based on their use of jargon who are worse at their jobs.
Dennis Rodman, The Anti-Hack
If you have a tattoo it is probably because of Dennis Rodman. Dennis Rodman was upstream of Allen Iverson who popularised tattoos in the NBA and the NBA popularised tattoos in professional sports which popularised and normalised tattoos for all.
In the 90s though, Rodman was 'Bad as I Wanna Be' a strange and unprofessional athlete who dyed his hair, wore a wedding dress and married himself, was routinely ejected from games and threw so many of his basketball jerseys into the crowd that his teams began giving him the imitation rather than official jerseys to wear in the games to save some $$$.
Yet Rodman is a 5 x NBA champion, 2 x NBA Defensive Player of the Year and 7 x NBA Rebounding champion. In fact Rodman is probably the greatest basketball rebounder ever. He averaged 17 to 18 rebounds a game in a league for which often the rebounding leader most seasons pulls down 11 or 12 boards. Which is to say, Rodman at his peak was around 60% better than the people we routinely regard as the best that year.
Rodman was a rebounding savant as well, outperforming much taller athletes in NBA centers as a power forward. He could score, but chose not to. Phil Jackson described him as somehow a 'selfish defender'
You look at Rodman, and he seems frankly, mentally ill. Which he almost certainly is. But what he isn't is a hack. Here is a video where he explains the famous triangle offense used by Phil Jackson to win 11 NBA championships with the bulls and lakers. In Jackson's coaching biography 'spiritual hoops' he describes that Rodman understood the triangle offense and how he'd fit into it, before he was traded to the Bulls.
Productivity
So something hacks be saying is "I've used "AI" to increase my productivity 3 fold/10 fold/50 fold." A computer scientist I know was relating one such story, where his interlocuter was telling him the exponential growth in his productivity.
Let us take a brief pause to shout "hallelujah!" "AI" has finally solved the productivity crisis. I guess even a small number of enthusiastic soloprenuers adopting "AI" early can boost my countries GDP if not into the double digits, at least a high single, like an 8 or a 9, after decades of nigh economic stagnation...
Anyway, pause over, when this computer scientist heard "50 fold" his bullshit detector started to go off, his critical faculties, that is to say intelligence began to kick in. He asked how this booster was measuring said productivity, and what the booster was working on.
Turned out this augmented "AI" cyborg with 50x productivity was out-of-work and had volunteered to do some basic development work for a public institution. They'd been doing that volunteer project for over a week now, but the computer scientist reported that the project was something that could be done pre-"AI" revolution, in an afternoon, or a day, in terms of novelty and scope.
I have personally heard similar claims of "boosted productivity" people stacking agents upon agents upon agents. Yet, if a pastry chef told me that their new kneading machine had boosted their productivity 10x, it would be unambiguous as a claim to determine the truth of, I would see 10x the pastries. Or the pastry chef working 1/10th the hours. Or some combination of increased output, decreased input.
I see none of this, and my personal suspicion, is that a bunch of hacks when they say they've "increased productivity nx..." are actually just counting the number of "AI" "agents" for example, if they have an agent that sends pro-forma birthday messages to their facebook friends, that means, to a hack, that they've just 2xed their productivity.
But I'm hoping to become a trainer of young business minds, and I've been reflecting on how I would best introduce this concept of productivity. Here's what I think I would do.
What can we produce?
Then I'd turn to the whiteboard with the marker that has 2xed my productivity, to note down goods and services...
If I don't get the contribution I'm looking for, I'll suggest it rhetorically:
What about waste?
Then write down waste. We produce goods, we produce services, and we produce waste. We produce a fucktonne of waste. We are extremely inefficient.
With respect to "AI" "art" the receipts are in - "AI" "art" is on the forefront of dead-internet theory. 3 years ago I watched "content" proliferate on a 16 year old's artstation profile where they "produced" a portfolio of "AI" images that numbered around 1000 pieces in a month, mostly of actress Emma Watson as Spider-Gwen from the Spiderverse with visible camel toe.
My post from three years ago was I feel impressively prescient in calling the likely impact on the internet art community "spam", granted I could not predict the specific jargon of "AI Slop"
But importantly, not only is slop (which does not deserve scare quotes) itself waste produce, it is itself produced wastefully in terms of resources like finance, land, energy, water and compute. I will link to a debunking of the more sensational claims, but much of it is real in terms of opportunity costs - like the money that has been invested in building ChatGPT5 could have paid off the student debt of the US, or the universal healthcare of US citizens.
But other things like, because there is just collossal demand for cutting edge GPUs, production capacity has been reallocated from gaming GPUs to datacenter GPUs. Those GPUs run constantly in the data centers and run hot and the GPUs will be superseded or fail within a matter of years. At the same time, what Deepseek outlined as a possibility, is that "AI" models can be used to train other "AI" models that hit the same benchmarks without any of the investment of the upstream "AI".
My point is, the waste, the production of waste. This is where I get a sick, sexual pleasure, from the way people stop in their tracks when I ask them "what is the economy for?" Your modal person has successfully been indoctrinated to believe in productivity. Even though economics 101 will point out that an oil spill - a massive costly environmental disaster - will be positive for GDP growth, because there is heaps of consumption.
Eating one cheesecake contributes to GDP, eating two cheesecakes doubles your contribution to GDP.
Most of the rich developed nations have been struggling since the onset of neoliberalism with slowing economic growth. The elephant curve is a piece of economic esoterica that is hard for laypeople, in which I include myself, to read. As explained to me, it basically says that the incomes of the 1% have grown massively, and the incomes of the middle class have been transferred to the middle poor, which is to say, people who lived in wealthy economies in 1980, all the GDP growth has not resulted in increased incomes for them, it was transferred to the BEMs - predominantly all the people China lifted out of poverty, India to a lesser extent, and probably Brazil to an even lesser extent.
"AI" aside, if you live in the Anglosphere, in Europe, in Japan, it is important you understand what your hard work has been doing for 40 years. Your hard work has been dedicated to giving wealthier people more wealth which they use to buy up more of the capital base, so they can extract wealth from hard workers even faster. This is called "growing inequality" most often.
The 4 Income Household
...Is not a sharehouse with four people in their mid-20s working minimum wage jobs while they finish accumulating student debts over 7 years. No. It is where the story I am going to tell about the right Productivity Crisis begins.
Alas, it is a story I need to tell backwards. And it is going to be a very reductive story, because I don't want to get into the weeds, so I won't talk about how inflation is calculated, pareto efficiency or the elephant curve. I can only hope that you entertain the story, and maybe even seize the opportunity to ask an economist about my story and see what they say.
1. In the 'Post-war' period if an employer grew its revenue from $100 to $120 what could happen was that the employer's profits grew by $10 and the workers salaries grew by $10. In the simplest terms, you could gain access to a share of economic growth through waged employment, referred to 'growth in real wages'
2. In the Neoliberal era (approx. post 1980) if an employer grows its revenue from $100 to $120 what happens is that the employer's profits grow by $20, and the workers salaries stay the same.
3. Stagnation in real (adjusted for inflation) wage growth can be offset by speculating on real estate, in short you use your steady income stream to secure a loan and use it to buy an "asset" or "investment" property, hoping your "capital" increases in price faster than wages making your "richer"
4. Because buying existing housing isn't productive, FOR house prices to rise faster than rents which in turn rise faster than wages (eg. House prices increase by 14% in 2008, while rents increased by 6% and wages increase by 2.1% just below inflation of 2.6%) it has to be fuelled by debt, specifically the house prices rise by the new owner borrowing even more money than the seller borrowed to buy the same asset. So in 1980 buying a property might have been at a price 4x the median household income, in 2026 it is 16x the median household income.
5. What you can borrow remains a function of what you can earn, and particularly how you earn it. Namely it is better if you are a permanent employee in a full-time job. By 2026 you may need to be a couple both in permanent full time employment, with parents who are able to gift you your deposit, as the deposit may require a young professional paying rent, as long to save up a deposit as people once spent paying off their mortgage. Those numbers are getting into the weeds.
6. As such, because of the inflated prices (and subsequent costs) of housing, for many people if their actual productive workload is less than full time hours, this poses an existential threat. eg. you can get done the work you need to do to fulfill your role in the firm's production of goods and services in 3 hours per day, but you need to earn the salary of somebody employed for 8 hours a day.
7. Such gaps can be bridged by 'pseudo-activity' where there is some degree of surveillance, supervision or scrutiny. Psuedo-activity can be accomplished through benign means by having meetings that are functionally social hangs, but from outside the meeting room look like a bunch of people in office attire discussing business, or corresponding through email stuff that is of no consequence but from the outside looks like an office-worker sitting at their desk doing a productive activity. It could also be accomplished through acts of self-sabotage for which I will link to a video to describe. But actually frustrating the process of individuals, teams, departments and firms accomplishing their goals.
8. Where there is some absence of surveillance, supervision or scrutiny, such gaps can be bridged by "gaming" the proxy measures of productivity or business. Which is to say, an employee simply needs to get good at making a 40 hour work week that is mostly taken up by doomscrolling, checking sports scores, looking up highschool crushes on facebook, playing solitaire etc. look like it is actually filled with work stuff. So an employee needs to master fabricating reports that make them look busy and important.
9. If the employee is at a remove, such as in a work-from-home post covid setup, this reporting can become truly trivial, and the activities they actually spend time on during their paid hours can graduate from things you can do in an office environment, to things you can do on a weekend - like walk your dog, watch movies and tv series, play video games, bake sourdough etc.
Sidenote: It is important to be clear that thru steps 6-9 if this story is taking place, it results in no loss of actual productivity in the economy. The issue is one of pricing, and marketing as in people are buying and selling 40 hour thru to 60 hour (or the Chinese 996 (?) 72 hour) workweek, when in many places a productivity audit would show only about 16 hours are justified.
10. Under the recent developments of wfh, somebody with an abundance of unscrutinised unproductive hours, can opt to rededicate those hours not to leisure, but to another job. So you can have a fulltime job as payroll clerk for an office (for example) by working from home, then do gig work video editing while your employer is paying you to be a payroll clerk. But why would anyone do gig work instead of baking, watching movies and walking their dog? Well because that extra income can reduce the burden of your mortgage and give you more disposable income.
11. WHICH if your mortgage once ate up 60% of your household income, and working two jobs simultaneously brings it down to 50%, THEN you have an edge in the housing market. You can buy in at a higher price to get BACK to 60% of your household income by taking on larger mortgages. Just as in 1950 a blue collar factory worker as the single income of the household could get a 20 year mortgage to buy their home, but by 2010 it was more or less obligatory for households to have two incomes in order to secure a 40 year mortgage, it is most likely that if FOUR income households become possible, they will over the course of a decade or so, become MANDATORY unless something disrupts the widespread speculation on existing housing stock in Australia, the UK, the US, Canada, the EU, New Zealand...what is often lost in housing affordability discourse, is that it is happening almosteverywhere. Argentina has a housing affordability crisis, the Phillipines has a housing affordability crisis. Russia does not! I'm going to hazard a guess that that might be because Russia is currently feeding it's economy into the misadventure-meatgrinder that is its invasion of Ukraine.
12. The main counterargument against the 4-income-households of the future, is largely because it is based on secrecy, but this is just a question mark. On the one hand, it could play out that the more people who double dip on their fulltime hours by working two-jobs simultaneously, the more people get busted doing it and lose one or both incomes as a result, until the practice is seen as 'not worth it' and dies back to a natural sustainable frequency, like most fraud and freeloading.
On the other hand, the practice could become so widespread as to become a new-normal where just as getting a sleeve tattoo, or even facial tattoo no-longer renders people unemployable because it is so normal, at some point it may become understood that the job market now consists of 'payroll clerks as a service' where they no longer pay a wage, but a subscription to a sole trader that is nominally discounted from the wage they used to pay a low-level accountant but they understand now that they've outsourced that function to a sole contractor who is the payroll clerk for eight subscribers they work for simultaneously. Much as businesses used to employ cleaners and now hire a cleaning service from a separate business entity.
Overemployment
I've been bullish on AIs ability to transform productivity pretty much from the get-go because of the taboo subject that we don't talk about how unproductive and inefficient our economy is.
It is one thing to replace a farmer's ox with a tractor and watch them plow a field in an hour that once took them all day. It is another thing to instruct an employee that mostly chases down co-workers who haven't put in their footy tips for the latest round to explore ways to implement "AI" to gain efficiencies.
What is important to note, is that many workers do work extremely hard. Hospitality workers get "slammed" by busy shifts, having to rush about completing so many tasks to a meaningful standard that they don't even get time to look at a clock. Telemarketers in a call centre work extremely hard, chasing often unrealistic sales quotas at a pace set by an auto-dialer while being emotionally drained by having to interact with so many strangers in such adverse circumstances. Drivers work hard. Gig workers work hard. Doctors and nurses work hard.
This concession was determined fairly quickly and in recent history around the world, thanks to Covid lockdowns. Almost overnight, a taboo was lifted and lists were produced labelled "essential workers" and there was an interesting pattern - work that is essential is more often than not, poorly paid and unprofitable. It also tends to be actually hard, in being physically and emotionally demanding.
Conversely, there are whole industries that are pretty much waste, they have long been known to contribute little, no, or negative value. Ones that come to mind are management consulting, which has an appalling reputation, and fund managers who famously underperform index funds.
I have had friends whose job it was to watch reality TV shows about real-estate, and then write summaries of the episode to post on a magazine's website as content. I have held a job for a period of years where for 6 weeks of the year my team was understaffed, for the other 46 weeks of the year, we got through our daily workload by 10.30 AM at the latest. This was one of the best organisations I have ever worked for, and our marketing department blew an entire marketing budget for a year on an ad-campaign that promoted features, rather than benefits, and allowed a competitor to run a campaign where they positioned themselves as having our product-lines actual competitive advantage. The manager of that department was eventually fired, some five years later, but was known for browsing real-estate websites all day, every day, all year.
We had a tech-support full time employee on a team of 4 that 100% of the time we put a customer with a technical inquiry through to, called us back to inform us he couldn't help the person, and that he was putting their call back through to us in the sales department. Before we could say in alarm "why the fuck would you put them back through to us?" the phone went boop and we were left in the embarrassing position of having to explain that if our tech support employee couldn't help them, then we certainly couldn't, and then we put them through to the guy in tech support that actually did his job via direct extension which we weren't supposed to do.
(I am actually fairly certain that this useless employee now lives somewhere in my proximity. I don't know if he recognizes me, I recognize him, but my reading of our crossings is that he would rather not speak to me nor acknowledge me, I fear he wound up unemployable and possibly alone. He could be happily retired, but then I'd feel he'd stop and chat to me...)
Okay, so I know this will be hard to follow, but as of this line I am picking up where I left this draft sitting for two or three months. Sorry if it's jarring - but productivity of employees is not a problem that needs solving when people are already overemployed, that we have many effective private-welfare kings and queens and inbetweens, we have many people whose wage is paid on the grounds of them being officially employed, but is not earned on the basis of doing productive work.
Their role exists out of habit, and has possibly expanded beyond the 37 hrs originally made available because they didn't have any basis for competing apart from outworking your competition, which is hard when there is nothing productive to do so instead, you adopt what Cal Newport calls "psuedo activity/productivity" essentially you look busy packing your day with emails, meetings etc. so you can leave what little work you actually have to do between the hours of 6am-9am or 5pm-8pm.
For many workers in this private-welfare situation, "AI" is the opposite of what they need. Because they are mostly occupied by trying to look productive without being productive. If "AI" reduces the 3-hour workload they have to get through in an 8 hr day, and can summarise meetings etc. but let me not overcomplicate this, if it can reduce a 3 hr workload to 15 minutes, that's really bad. You now need psuedo-activity not to get you through an 8hr day so you can do your work outside of official time, but you need pseudo-activity to take you all the way from 9am-7.45pm.
Now there are certainly roles out there that are tremendously productive - GPs have to see patients every 15 minutes of their working day, anyone on a casual wage has to produce for every hour or minute they are paid. They muddy the picture and also have the nasty habit of projecting that "most" people work as hard as they do.
There is just too much testimony as to the phenomena of Bullshit Jobs. In the first instance to explain the virality of Graeber's article when it was first published, to reconcile it with a world characterised by efficient markets, you would have to test something like an 'envy' hypothesis in contrast to the simpler explanation or 'resonance' hypothesis. The envy hypothesis would suggest that disenfranchised people took to the article because it reflected how they want the world to be - motivated reasoning to say that the game was rigged. Plausible, but assumes only a negligible fraction either have a bullshit job or have encountered someone that appeared to have a bullshit job.
The resonance hypothesis assumes the opposite - that actually a large number of people will willingly testify as to having a bullshit job.
Now, I am open to research into the topic nullifying the existence of bullshit jobs, but I personally would still require an explanation as to why I have been so overemployed so many times in my life and all the incredibly unproductive people I've observed simply hanging around an office all day.
A 2021 study claimed to debunk Graeber's claims, especially under the provision of Graeber's that one should only label one's own job a 'bullshit job' and never force the label on another.
I do however have a problem methodologically with this line from the wikipedia summary, that I find telling:
bullshit "taskmasters" and "goons" such as hedge-fund managers or lobbyists reported that they were vastly satisfied with their work, while essential workers like refuse collectors and cleaners often felt their jobs were useless.[17]
That might debunk Graeber's notion of Bullshit Jobs as essentially a subjective state. But intrinsically we still have a problem here - a campaign like a "day of absence" or even a good old fashioned NYC garbage strike should disabuse refuse collectors and cleaners of any perception that their jobs are useless. I am surprised to the point of suspicion, unless the EU with it's greater welfare safety nets just indicate that cleaners and refuse collectors feel their jobs should be replaced with robots already (more socialist economies adopt automation much faster and with less resistance than more neo-liberal economies as a general rule.)
A 2023 study titled "‘Bullshit’ After All? Why People Consider Their Jobs Socially Useless" used data from the RAND thinktank's longitudinal study on workers, and concluded the data supported a robust finding in line with Graeber's notions. The paper itself is fairly brief and fairly accessible and also refers to the earlier 2021 Sofia et al. study.
 |
| https://journals.sagepub.com/doi/full/10.1177/09500170231175771 |
You might have to blow that graph up, however something I find interesting is the top three % wise occupational fields that rank highest for self reported 'socially useless jobs' being Transportation and material moving, production occupations and sales and related occupations - the last of which the researchers classified under 'not only useless but actively harmful to society' but the first two you'd think COVID lockdowns established quickly as essential occupations.
What's intriguing, and I can't prove, is the connotations of consumerism. Because a truck driver delivering fresh produce to market (or supermarkets) I wouldn't think would self-report their job as not needing to exist. Fill that truck with idk...pokemon cards, labubus, halloween decorations, smart watches...anything from Temu, Shien, Amazon etc. and yeah, it becomes more understandable that if you knew you were shipping, for lack of a better catch all term - consumer crap around, I would guess being in transport would feel pretty worthless.
(This was a sentiment I heard fairly often expressed by marketing students who wanted to change degrees or retrain - that they had no interest convincing people to buy shit they don't need)
From there it is not too far of a step to imagine a bunch of people in production jobs aren't manufacturing cars, but like Xmas ornaments, corporate merch, plastic plates etc. So even though you are making a tangible good, with such repetitive work of course you are going to ruminate on how nothing would happen to society if you stopped manufacturing crap tomorrow.
Ed Zitron refers a lot to "the business idiot" which is close conceptually, if not identical, to the professional managerial class. Something like that, I suggest, was persuaded "AI" had a real problem to solve, because not only do they believe their organisations consist of passionate, efficient hard-workers, but that's what all their subordinates tell them too.
People who occupy a position in our society, that are poorly positioned to notice the prevelance of people who are employed far beyond any need for them to be productive. As such, they were blind in 2023 to something that was quite obvious to many, between 20-40% of the workforce roughly, that when so many people are at the office for 8 hours, to do 3 hours of work, there is nothing real to be gained from integrating "AI" into our workflow.
Dick Dawkin's has never been taken down so well...
Again apologies that I am rushing this to publication, I really wanted to try and find a source for this video's claim that when literacy was introduced many people that that books were talking to them. Books! Fantastic if true, but I haven't been able to verify this claim on history.
It's a really great takedown, to the point of burying the lede.
The Rushed Conclusion
So if the arse falls out the bottom of the "AI" bubble, this will largely be a financial matter that I anticipate will also have a chilling effect on boosting "AI" as the future of work and shit.
As the chips lay, there's really only SpaceX as a listed company that has 98% of it's valuation based on "AI". While I write this Anthropic has not gone public nor is there any suggestion OpenAI can. What likely will happen is some sizeable drop in the prices of their largest customers like Microsoft and Google (alphabet) as their "AI" holdings are reevaluated.
For those two US frontier firms, a $2T valuation will likely become impossible, but the real devestation is that "AI" is the US' growth story. Quite literally everything is banking on it, so like a GFC 2008 style financial collapse, wealth destruction.
Ed Zitron's interview with Jon Stewart is a better source to explain why Datacenter's are unlikely to be the bubble surviving useful infrastructure that fibreoptic cable was in the dot-com crash.
What I imagine however, is how the hacks will react. Hack office workers, hack linkedin posters, hack consultants and hack journalists.
The underlying thing being, if this technology was truly useful, it would not be hard to find a way to pay for it. People pay for value, and with these LLMs the value is just not there. There is no faster way to not be taken seriously as an artist than by producing 1,000 pieces for your portfolio in a month. There's no faster way to lose your reader's attention than to write something with "AI", there's no way to tell if your career advances if it's because you adopted and integrated "AI" or whether its because you adopted and integrated the latest corporate jargon which happens to also be "AI".
If these companies fold, if the largest infrastructure build out in history goes bankrupt, then I imagine the day after on that zoom or teams call and I can't imagine the hacks showing contrition. They can't show contrition because they will be convinced that they couldn't have possibly known that "AI" was a bust.
And it's true, they couldn't possibly know that "AI" was a bust, because they were hacks.
That will be awkward.
The tide will have gone out and we will be painfully aware of who is naked, and I just have this deep wrenching in my gut, that such a downturn will result in layoffs and that the hacks will keep their jobs because they are so often middle management, business idiots, and the competent will be ousted.
No comments:
Post a Comment