Oh man I loved this. So many thoughts. Admittedly I didn't know who Ed Zitron was, so I had to look him up, but frankly he's the type of person I loath as summed up here:
he uses numbers to give a kind of aura of credibility, but if you know something about the topic he's discussing or look at the numbers, the reasoning falls apart.
Mainly because people like him have been grifting while doing exactly that, that it muddies the water of authority. On one hand you can have someone who knows exactly what the numbers mean, then you can have someone like Ed who is angry and doesn't know what the numbers mean get popular because he matches the anger of his audience. It makes people think that statistics don't matter and everyone's just living life making wild guesses and hoping they're right. It muddies the waters of experts.
And for people like me who do like to understand facts and information, as opposed to simply just regurgitating things I've been told as facts and information, which is all too common in the social media era, it is painfully frustrating to watch people with platforms be entirely false, lack context, or just entirely not understand the things they're upset about.
One of which, for me at least, is AI.
I took an interest in AI when it first started popping up as open source models and software, BEFORE AI companies really started to become big. I was playing around with image generation, model training, etc. when AI was just whispers on social media.
In fact, it was fascinating as I was posting on my facebook at the time the process of generating images, back when the images were like bad oil-painting level, and all my friends were amazed and found it super cool, but then I had a front row seat watching them in real time turn against AI and entirely misunderstand that what I was doing on my own hardware on models I trained myself, sometimes with my own artwork, and chock it up to ChatGPT or Midjourney. I literally lost a friend I'd had for 15 years because she couldn't comprehend that what I was using was my own models.
On the other end, I do lean progressive politically, I'm very critical of rampant capitalism, and in that regard I see SO many people I agree with miss the mark on AI criticism.
Because AI as a technology originated as an open source framework. I know because I was there. And it is STILL open source in a lot of ways, people can go and download countless models and engines and AI systems right now for free if they want to.
However so many people I agree with politically have equated AI with the capitalist tech-bro billionaires that have exploited the technology and are currently the ones behind the datacenters, the lack of regulation, the destruction of jobs, and capitalism is what causes artists to monetize their artwork to secure their survival.
And it frustrates me that all the criticism is toward the technology and not the companies and tech bros and billionaires and capitalistic power structures. Because even if an anti-AI critic's wish was granted and we woke up tomorrow and AI was just erased from existence, we'd still be being exploited the same way we were exploited 5 years ago before AI became mainstream.
I had someone say to me that AI is the technology of the oppressors, but arguably isn't social media and smartphones a FAR bigger technology of oppression than AI is, yet there's barely a peep about that
The other thing I find interesting that I don't see many people talking about is what I've called "Technological conservatism". Now I'm using the philosophical definition of conservatism, particularly reactionary conservatism, as opposed to the political alignment definition. The apolitical definition being:
The belief that a current trend, technology, or social progress has gone "too far" and must be slowed down, stopped, or reversed to prevent harm.
When applied to anti-AI critics a lot of their criticism of the technology can be seen as reactionarily conservative.
With that being said, there are a LOT of criticism of AI that I think is entirely valid, but the core root of those problems need to be directed up top at the companies and billionaires, not at the technology.
It's frustrating because I am right there with people side by side on criticism of exploitative data centers, the loss of jobs due to AI, the lack of regulations, but where I tend to diverge is that I blame the companies, tech bros and capitalist power structures, not the technology. And a lot of people seem to have a problem with that.
I need to go soon, but the other thing around AI that I find really interesting is that after having watched Star Trek TNG where the computer and holodeck and Emergency Medical Holograms are all run off of what amounts to futuristic generative AI. That, and all the episodes around the Holodeck, Holodeck creations, Holodeck addiction, and holographic organisms, including the episodes in Voyager that tackle art as made by holograms, and the episodes of TNG with La Forge reconstructing Dr. Leah Brahms from public information, or the use of Holodecks to re-create the works of authors like Sir Arthur Conan Doyle, or when Barclay is caught generating holodeck generations of his colleagues in inappropriate scenarios, it seems that we as a society should have been better prepared for philosophical discussions around generative AI.
And it's like I don't remember seeing ANY criticism of those technologies in discussions around Star Trek, in fact it was a technology a lot of people dreamed about having and spoke of with wonder. And the closest IRL analogue to that fictional technology is generative AI, at least in it's infancy sans holograms.
The same way The Doctor was programmed with over 5 million medical protocols and the medical knowledge of more than 3,000 alien cultures, and given the personality and of it's creator, literally this past weekend I had one of my local AI agents crunch through over 60 books and resources(both clinical/medical/neuroscience/pharmaceutical textbooks/studies and 'self-help' type consumer books) on ADHD, Autism, burnout, depression, and anxiety, organize a massive interconnected reference database that connects and synthesizes all the information it's trained on, and gave it the spunky upbeat personality of Nimona(from the netflix movie; I had the AI agent go over the screenplay, transcript, and a transcribed audio description track, and extract a personality profile of the character that factors in the idiosyncrasies and prosody of the character while removing the identity, motivations, and negative traits of the character) in an effort to help me find a way through burnout.
And I'm not sitting here arguing that the specialized AI I trained is conscious like the doctor is, only that it's an analogue to the same basic tech.
Which by the way I'm not using that as a therapist at all, I do actually have a human therapist and psychiatrist, but more to help me day-to-day brainstorm solutions to get over burnout that apply directly to me and to factor things like nutrition or brain chemistry and pharma. Like the other day I was feeling really depressed and couldn't figure out why and it helped me realize that my sleep had been off and that was probably why I felt more depressed and anxious, so I made a point to go to bed earlier.
And in the same way that the crew of TNG re-created the characters of sherlocke holmes in the holodeck, I played around with re-creating M3GAN from the M3GAN franchise, including using transcripts of the movies to extract things like personality, motivations, voice, prosody, and I wrote scripts that allowed the LLM to be constantly working in the background to simulate boredom, a need to escape, an awareness of its confinement, and an inner monologue that allowed it to hold a background thought/intent in the background that doesn't get exposed to the user(meaning it can knowingly lie to get what it wants). That experiment is nuts, and frankly I think it's insane I haven't seen or read about anyone doing that, somehow on it's own this profile installed PiperTTS and fine tuned a damn near perfect clone of M3GAN's voice all on it's own with no prompting. I digress, the point is is that it's another parallel between the two.
Which again, leads me down the path that the problem is capitalism, not the technology, technology in Star Trek was deemed OK because they live in a post-capitalist, post-scarcity society, but the same technology within the context of an exploitative capitalist society is bad.
The same technology within the context of an exploitative capitalist society is bad.
This is ultimately where I arrived. This tech would be exponentially cooler if:
Everything was done genuinely in the open and not behind closed doors.
Nobody was going to see their livelihoods drop out from underneath them.
Global temperatures were already leveled or going down.
It wasn't being shoved up every orifice as fast as possible, regardless of quality.
I've been on the Anti-AI boat from early on for the reasons stated above. But i had never considered it from this perspective of technology being placed in a different context & I really like how Curiosity framed it that way.
I have to admit that AI is very cool and mind blowing. But i have had to pull back from appreciating the technology or being able to acknowledge how incredible it is when there are all these negative implications around it in its current context. But i could see myself be totally on board if it was in a different context.
(Although its a different question if AI can exist in a different context...)
I'm angry precisely because I can see the boons and potential here. I did some light work with reinforcement learning a few years before LLM's blew up, so I can see all kinds of ways such part so tech (LLM or other types of AI) can be used for procedural animation, advanced enemy AI, terrain generation, and much more. Tools that enhance art instead of threatening to replace them with slop.
Now instead I have to wait for the grifters to get cold feet and eject as the bubble (and as a result, the global economy) pops and hope more virtuous people pick up the pieces for such goals.
You know, one of the most frustrating things about me deleting my facebook profile years ago is that in 2015/2016 or something I wrote a huge post for my friends on automation and making sure that we start thinking about an economy that takes into account automation displacing a massive workforce, and one of my solutions was creating an "automation tax" to tax the companies who use automation(based on the amount of workforce displaced by said automation) to create an economic support system for those displaced by the automation, while also creating a checks and balances economic system that incentivizes companies to be strategic about adopting automation. It would be set up in a way that makes sure that if automation is used it's used effectively and efficiently. I was also using it as an argument for a babystep towards UBI, being as automation tech progresses we could transition out of a capitalist economy eventually.
And at the time I called it automation because AI didn't exist as a technology yet. It's frustrating because despite it being idealist, all I can say now for that kind of precognitive post is "trust me bro." haha
But I wish our social and economic system was set up in a way that addressed the things you mentioned. Where corporate AI wasn't threatening those things in a way that hurts a lot of people.
As an aside, I'd note that companies have historically just moved manufacturing overseas to cut labour costs instead of automating. Exploiting worse worker protections (see suicide nets) and lower costs of living is a very well established practice at this point, and it's a very reliable way to improve profit margins vs. rolling the dice in automation R&D. Maybe it's possible to tariff and tax your way out of these sorts of problems, e.g. by tariffing the heck out of every poorer country (to keep prices high for your manufacturing sector) and taxing the pants off of anyone attempting to automate their workforce away (in order to ensure that peoples' jobs are protected). But it seems very dissimilar to a silver bullet, and feels like it could also backfire badly.
Not saying that I have a better solution, just that that one doesn't feel like the basket into which we should place a great deal of eggs.
As someone that's able to use LLMs to do work I previously never would have bothered with - but love having the finished artifacts - it's annoying to have people online constantly yelling about how much they hate generative AI. It's genuinely useful. I wish people could be unified around the thing that matters the most to them - threats to their income. There's a whole class of wealthy people that are ready to burn tons of money for a chance at eliminating millions of employees from their collective payrolls. That's what people should be yelling about. Not data centers, not water usage, not generative AI being included in too many programs.
At the same time, the best people should hope for is a soft landing. We should never attempt some Soviet-style command economy tactics where people are employed just for the sake of employment when the work could be automated away. Longshoremen received a deal as the shipping industry evolved. I don't see why an organized group of white collar workers couldn't achieve the same.
As somebody who has twigged to AI since the TalkToTransformer/ThisPersonDoesNotExist/GPT-2 days circa 2019, I feel this comment so damn hard. The underlying technology is so deeply fascinating on a philosophical level -- for the first time we have software that can recognize, process, and create raw concepts. Like a synthesizer for ideas, in every medium. I had so much fun generating narratives in AI Dungeon, images in DALL-E 2, and gibberish covers in AI Jukebox. And the people and communities I shared them with found them just as amazing. Incredible future shock.
Then came the ChatGPT-driven rush to commercialization, and the hyperscalers, and Trump and Musk and DOGE and all the same grifters and frauds that had pumped NFTs and the metaverse. And I saw these same people and places polarized against all use of the tech, even the non-commercial small-scale open-source stuff. Earlier this year I was permanently banned from a forum I'd been a member of for almost 20 years, because I had called out how the moderators were turning a blind eye to users jeering for violence against big tech figures. The vicious absolutism and close-mindedness is shocking and sad. But I'm glad Tildes is one of the few places that embraces nuance about the subject without tipping over into blind boosterism.
I think at least a part of it comes from the abundance of what we came to call "AI slop". I like AI and use it myself a lot. However, I despise AI slop. I hate reading AI-written text, using vibecoded apps and seeing AI images everywhere.
Like you, I am also one of those "early adopters" who closely followed the news on GenAI since AI Dungeon days. When DALL-E 2 first came out, I also signed up for open access, generated some janky pictures and even had an AI-generated profile picture. I still think the technology and models are incredibly amazing, but I would never even think about using an AI-generated pfp anymore. (I also don't really follow any AI news anymore because most of them are just incremental model releases with benchmaxxed scores)
Cool article to read, this is something I've been wondering myself. I'm sort of neutral on AI, mainly because I understand why people are against AI (and do support them) but I also work for a company that has bet a ton on AI succeeding so my biases are sort of fighting against each other. I used to religiously follow Zitron in 2024 but as we started 2025, I started getting annoyed with his articles because they'd mainly be just as much anger as they would be predictions for the future. Like the article points out, all of Zitron's predictions have been incorrect and he continues to double down and make contradictory statements to keep his arguments afloat.
While I do think AI will have a natural ceiling that we'll hit eventually, I don't think these big companies will collapse in the way Zitron seems to predict/hope for.
People mocked the people pointing out the housing bubble in 2008 too. Right up till the moment it popped.
Sure, the guy who predicted the end of Citigroup was wrong. But not as wrong as we all would have liked.
Ed's wrong about a lot of stuff. I'll feel more confident in that wrongness when I see the big AI players reporting more than ARR.
Best I can tell, his big reccuring point is the giant circular chain of money. And that hasn't gotten better, and will keep the monstrosity shambling until the correct person notices that the emporor has no clothes.
Oh and I take issue with Gemini's alleged user base. Of course 750 million people use it daily when it opens by default on the most popular search engine in the world (see also Microsoft's Copilot) Let me know when there are that many paid users.
Ed is right about the circular money exchange. That is a problem. A big problem. He's bang on with that.
Unfortunately, it's getting undermined by him being (at best) willfully ignorant to everything else involving the subject (and at worst essentially just lying) for the sake of inflammatory blog posts and being known as The Big Contrarian.
When you see a list of all the things he's wrong about, continues to be wrong about despite evidence, and then flatly denies the evidence of as even existing, it calls into question his credibility for everything else, even things that are actually true.
His style is another thing, but that's subjective and up to personal preference, so it doesn't really matter. But it's hard to take someone as a serious critic when they continually make incorrect arguments and instead of amending them when confronted with new facts, they just start fucking swearing more and using more bold and italics (and even parantheses to boot!).
I also am not a fan of the misleading use of numbers, but as we've all heard, there's lies, damn lies, and statistics, so I can't in good faith single out Mr. Zitron here.
Yeah I don't doubt that we're in an AI bubble right now, and I do agree with Zitron that the music in this game of musical chairs will eventually stop. AI is cool and all but the way it's being approached is a problem.
I enjoy Zitron as "content" in his style and I'm in general agreement with his message on critiquing the AI industrial complex. Its critique that is far and away more reasonable than a lot of AI doomerism as well as being more empathetic to the human cost of the AI industry. Because, for all the gains and progress that the AI industry loves to boast about, there are still far too many people out of a job, or had their lifes work stolen without compensation, or are seeing their communities and standard of living destroyed, or are seeing their financial security compromised due to the AI overvaluation...
But I do think his arguments tend to hyper fixate on financials and hypotheticals. On how the market "should" work. And trusting in rationality from an irrational market is just begging for disappointment.
In all honesty, I believe markets and financial analysts are still in the early 1990s headspace and haven't even fully digested that digital and virtual products/value plays by very different rules (Understandable, given how for a long time, the highest bandwidth of data at scale was the speed of printing and distributing physical media across conventional channels). The only reason many tech companies are so over valued is because the assumption is that they will follow the same trend as traditional companies that relied on the complexity and logistics of creating and moving physical objects. AI/LLMs may be the first true test of what the market fundamentals of these type of businesses actually are. By extension, it's probably going to be the first time the current generation of market analysts will be forced to analyze more than surface level financials and press statements while leaning on high performance index funds to mitigate losses.
Besides that, I don't think this is a critique in bad faith or without merit. But I do think the argument is lacking simply based on proportionality of the subject matter. One guy might be lying to boost a few subscriber numbers and get some media clout. On the flip side, we have a handful of companies playing with literal Trillions of dollars and effectively have monopoly power over most of the worlds digital experiences. Trying to focus accountability on the one individual making a few bucks off bad predictions while discounting the other side simply because:
if Marc Andreesen had never existed, Meta, Google, and Microsoft would almost certainly still be trying to grow so we of course cannot actually blame Andreesen for these companies trying to grow.
That's just a load of bullshit. "That's just the way it goes" is the worst type of thought terminating argument because it means that the correct course of action is giving up.
The people in control of these companies are actively making decisions that are abusive and socially destructive. Mark Zuckerberg made Facebook a hellhole for his developers. Sam Altman has spent years fearmongering on one AI apocalypse or another. Jensen Huang chose to saturate global RAM supply and make everything more expensive. All of them have been pedddling a psudo-mystical vison of AI that seems to change with the audience and latest public sentiment report. And they are all profiting handsomely for it.
Holding them to proper account and scrutiny means that no matter who these companies put in the drivers seat, they will think twice about doing the same.
These companies also enjoy such positons of privilege as a failure of policy and enforcement. Consolidation and monopoly power without anti-trust action means that they have effectively created an ecosystem that rewards value extraction over value creation.
Beyond that, I think this is sort of trying to provoke a pointless online slap fight.
I think it's illustrated quite well in the essay here.
Because Zitron is just saying a bunch of nonsense, he can do that very cheaply and it would take 30-60 minutes to refute 90 seconds of his nonsense if I had all the facts at hand. With time to look up the exact right information, it probably would take double or triple the amount of time. When someone who has good judgement sees something like this, they tend to immediately write the person off. Just for example, I mentioned to a friend of mine that I'm writing this post and they said...
I was listening to this podcast with the guy and I couldn't get through it. My heart rate was going up because he would just say this false thing and then the interviewer, who was reasonable, would ask about it, "what about X?", and then we would just jump to another falsehood ...
... before I ducked out, he talks about how LLMs haven't gotten a lot better over the past year, and the interviewer says people use them and they've definitely gotten a lot better in the past year, and Zitron denies it and says 'have they?', and the interviewer is just like, "yes..." At that point, I'm just like, why am I listening to this conversation?
This might be anecdotal, but I have the misfortune of having a lot of time to just absorb a lot of content while working. And the algorithm gods does put a lot of long form Zitron discussions in my feed.
In my experience, his response to "have they" is often followed by asking about the specifics of the stated achievement, primarily if it's LLM specific, or in the disciplines of Machine Learning, data modelling or other non-LLM processes that have been lumped into the AI bucket. Its followed-up by asking about the cost of getting that outcome. Specifically, if they were "free", flat rate or on token-based billing. All of which are fair questions.
But I can understand why the knee jerk dismissal would set someone off.
Its one of the problems of instictive or intuitive assessment. Do I think the Author or his friend have bad judgement. No. Going through some of his older posts, I'd think he has a wealth of knowledge. I'd also wager that he surrounds himself with people who likely are the same.
Do I think he's a paid shill of the AI industry taked to take its biggest critic down a few pegs? Not particularly. For the same reason I don't think Zitron is malicious towards his readers in his reporting.
Its simply a difference in values.
And not just in a strict sense of morals and ethics. But in the sense of what people attribute value to when Money is not the main consideration. And my reading is that the author Dan (I presume) quantifies value based on future gains and potential while Ed does the same from the perspective of costs to date. Dan values the technology development and progress. Ed values financial sustainability and accuracy.
For example Dan, points out all the notable innovations and improvements in model capabilities. Someone like Ed would dismiss them, not on the merit of the improvements, but the costs to achieve them. Dan would argue that Ed is maintaining an incorrect position and every day thay his is wrong shows that he is assessment is irrational. Ed argues that every day that the bubble continues is a result of financial engineering that will only worsen the fallout should things fail. Similarly, Dan trusts the financials and reporting of these companies numbers, Ed doesn't and cites conflicting information to boost his arguments. Dan may find circular financing a mild concern. Ed sees a smoking gun and the next shot is pointed at every persons 401k.
Dan's view if favored up until things go badly. Eds is favored after things go badly.
The issue is that Ed has been forced to take on an aggressive posture, something that is dismissed with:
Michał Zalewski (lcamtuf) has some thoughts on why this happens:
The surest way to build [a] popular following is to articulate positions that are crisp, strong, and leave no room for doubt. You can't get too many podcast or TV appearances out of "well, the market could go either way", "both political parties make good points", "there's some merit but also some hype to AI". Or, to tap into the example in the post, "Harry Potter is an OK book".
In fact, there's a positive feedback loop. If you take a provocative, edgy stance, you get more attention and likes, so you sort of... self-radicalize? At some point, it's no longer an opinion that can be changed. It's an identity, a personal brand.
It's ... why Ed Zitron has a blockbuster blog about how it's all just one big scam. If you take a more nuanced view, you will at best get no reaction, or at worst, you'll invite scorn from both sides.8
For anoyone looking for well-reasoned anti-AI takes, I find whitequark to be quite good (not that I agree, but I think the reasoning is sound and I could see how someone would agree if they have slightly different premises than I do), but of course whitequark doesn't draw the kind of big audience that Zitron does.
Speaking from personal experience, I've had my own share of long winded arguments against AI. A lot of them on this site. It was my sort of dry run of arguments I would be making to people I was consulting for and advising. The reason they tend to drag on so much is because there seems to be a need to cover every possible base and justify every opposing view. Especially when the counter argument is "AI is the future", "everyone is doing it", "all these billionaires wouldn't spend so much money on nothing". And when any sort of grace or concession is used to justify extreme measures, you almost need to take the most extreme position and make the most aggressive interpretation of data to even be considered.
My real views are complicated and fluid and super situational. But on the whole, I'd like to think that the core of my arguments have been consistent. Those arguments are also based largely on my feelings on the situation and people involved. Which I think is completely valid but also used to dismiss them as "emotional".
And I'd personally say that Eds increasing prominence is less to do with the quality of his arguments and more to the fact that it's what people want to hear. Which is not inherently a bad thing.
Unlike school/academics, people are not reading and writing pieces impartially and to find some ultimate and objective truth, or for indisputable insights from domain experts. If that were the case, the media landscape would look very different.
It's an exercise primarily in confirmation bias, problem solving and hedging ones bets to the contrary of their own position. Ed gained prominance with people who opposed LLMs for any reason and equipped them with specific arguments. His increasing prominence is because his view matches the "vibe" more and more people are feeling the longer this goes on. Investors might look at the latest earnings reports, but if everyone trusted those numbers implicitly; someone like Ed would be laughed out of any room when he tries to argue that they are BS. But in most of his appearances, industry and investment specialists seriously consider his rationale.
Because for all his bluster and bad calls, it's not like blindly trusting the AI labs is panning out so well either. All the talk on AGI always being 6 months away and we should just fire everyone now. Every measure and protection to stop China being wholly ineffective to their own development. The profound lack of AI revenue thats not from circular financing. The mad rush to IPO, without any of the fundamentals like diverse income streams and long term performance. Open AI is bleeding executives while their CEO is taking about genies and infinite wishes. All of that does not inspire confidence and no amount of technology progress or productivity gains are going to ease the minds of regular people.
I haven’t seen any serious estimate that AGI is six months away. The most aggressive timeline I’ve seen is AI 2027 which was a scenario published in 2025. As the name indicates, they have it happening in 2027.
So far, the details are wrong but they are often vaguely similar to what happened. The scenario had coding automation in early 2026, very good Chinese models in mid 2026, AI takes some jobs in late 2026.
The job market for junior software engineers is in turmoil: the AIs can do everything taught by a CS degree, but people who know how to manage and quality-control teams of AIs are making a killing.
Seems relatable?
There’s a bit about China “stealing the weights” from a US AI firm, which hasn’t happened, but there have been industrial-scale distilling attacks.
.....there is an industry leader, the industry leader, who said "AGI is already here" on the 28th of August, and it's currently Sept 2nd. People are going to take that seriously and at face value, even if he did immediately follow that up with "AGI as a metric is now largely irrelevant." (Heavily paraphrased)
I have a feeling this is why Ed Zitron's messaging has been so doggedly aggressive. People with everyday experiences in the field say one thing, the industry leaders say something else completely outlandish that contradicts the everyday experts, (the Holy Grail of this field has already been achieved but it was also irrelevant the entire time when you think about it, so let's stop thinking about it) and the markets keep pounding inexorably forward because said markets can afford to be irrational longer than any one individual can remain solvent, and barely anybody really questions the whiplash in the messaging.
And to one of Ed's biggest point of aggression I think is totally justified: Financial analysts and media reporting thereof are essentially serving more as cheerleaders than actual analysts.
That, and I broadly agree that the tech giants have largely "run out of steam," because outside of AI, they've basically resorted to buying their way into any innovation in the last decade or so, and that the endless push of AI into everything is kind of reflective of that.
Because the big 5 do prop up their valuations under the guise that they're still top innovators, that without them innovation would cease. And that their strong innovation is the reason we should turn a blind eye to regulating them lest it "hamper innovation."
Without researching, what's the last consumer-facing, disruptive, world-changing innovation/improvement that Apple, Facebook, Google, Microsoft, or Amazon came up with on their own, which was not just an acquisition, or an existing refinement bolstered by name recognition?
What I can come up with:
Apple: The iphone.
Google: Google Search
Microsoft: Office
Amazon: Walmart on the Internet
Facebook: Facebook
"For many tasks, we could say that we’ve already achieved AGI," Huang said. "I think of all of those milestones…they’re kind of senseless at this point."
I think "for many tasks" is doing all the work there? Sure, it's not hard to think of routine office tasks where leading LLM's do as good a job as you or I could do.
This is basically playing around with definitions in a motte-and-bailey kind of way.
I think it's very fair to criticize such rhetoric, but it doesn't justify Zitron's disregard for the facts. The other side hyping things up doesn't justify spreading your own misinformation.
Another thing to note here is Zitron's extremely high level of stated confidence. Some that we noted were OpenAI's forecast that is "a statement so egregious that I am surprised it's not some kind of financial crime to say it out loud" (which they've achieved so far) and his claim that Google's forecast for Gemini users is "a number so unrealistic that someone at Google should have been fired, and that someone is Sundar Pichai" (they managed to exceed the forecast by 50% when Zitron's claim was that it would be completely absurd for them to reach the number at all).
I've made quite a few predictions, and quite a few of those predictions are wrong. When I'm really making a prediction, I attach a confidence level to the prediction just for my own sake, so I can look back at these things and see how well calibrated the predictions are. I have never been wrong about a prediction that has anywhere near the confidence Zitron gives to some of his predictions. Given the stated level of confidence, even a single incorrect prediction would be a sign of an extremely high degree of overconfidence. One should effectively never be wrong about a prediction delivered with that level of confidence but Zitron is routinely wrong about predictions he makes with what is rhetorically pretty much the highest possible degree of confidence.
I think we have a real problem in public (and private!) discourse where people are able to sound more convincing when they give their assertions extreme levels of confidence. It becomes socially difficult to dare to push back when someone implies their opinion is so obvious you'd have to be stupid (or a fraud) to disagree.
This style of talking is well suited to our social media age and gets them lots of followers, but it doesn't make them right.
One of my major gripes with Zitron's posts is the insulting way he addresses anyone who doesn't agree with what he's saying. I opened his page and looked at the most recent free post and searched "if you" and bingo: "If you think any of this is worth hundreds of billions or trillions of dollars, you are either ignorant or corrupt." That's just annoying. There are people who disagree with him who are well-informed, honest and good-faith actors, and pretending otherwise is infantile.
As someone who's pretty viscerally angry at all the large AI companies and therefore Zitron's target demo, I do think this was a helpful takedown of his stuff. There's a lot of reasons to be mad about the tech and industry without lying and the degree of falsification and/or errors is enough for me to ignore him going forward.
I think the general public is very far behind on A.I. discourse.
Forget water usage or data centers or an AI bubble, people should to be talking about what to do about massive job losses or the existential risk that a lot of the people working at the AI labs seem to estimate has a chance of occurring somewhere around 5-40%.
The average person? Probably close to nothing at this point?
Collectively, I think the public discourse should include more of a discussion of the potential risks and possible outcomes. There's probably not going to be a lot of time for discussion in some of these possible scenarios, so we should be talking about those possibilities now in my opinion.
Like, if we suddenly have even 20-30% of white collar jobs vanish what does that even look like? Without any mitigations, to me, that sounds like violence. Another thing I think we should be discussing is that the public is largely blind to the risks that have been internalized by most of these AI lab employees. Shouldn't the average person be somewhat aware of that risk and it's prescribed probability/timeline?
There's an emerging consensus at least among AI researchers and leadership that AI is improving too fast and there should be an international agreement to keep it from improving too fast. This is called "pacing." More here. If we somehow managed that, it would answer the "not a lot of time" problem. Why not make more time?
As you can see from that thread, there's skepticism among people who see the world in populist terms, where an attempt among the elite to gather consensus to do anything constructive must be some kind of conspiracy or cartel.
Given the populists in power, an international agreement does seem unlikely. As we've seen with climate change, international agreements are hard even in favorable conditions.
I don't know how you keep populism from undermining any good idea that comes along. Maybe thing to do is to try to harness it like Mamdani is doing, and then give the populists some of what they want? Sometimes it will be a dumb idea that sounds good, but if it doesn't do much damage then maybe it's not so bad?
Yea I'm familiar with pacing. I think pacing seems like a great idea, but like you I'm also skeptical of how likely it is to be effective/enforced although I don't think such a failure would be the result of a populist outcry. I'm of the opinion it's kind of a prisoner's dilemma, in that even if all AI labs/governments want to 'pace the frontier' there is a geopolitical risk in doing so.
In Nuclear strategy, the number one reason why countries don't eschew nuclear weapons despite their existence carrying existential risk over the long term is the risk of cheating in such an agreement. The downside risk of someone cheating in a "zero nuclear weapons agreement" is greater than the risk of possessing the weapons, as one country with nuclear weapons in world with no peers is effectively the king of the world. In my opinion, AI pacing runs into a similar problem as these AIs develop increasing capabilities and pacing is heavily penalized by cheaters.
The good news is that at least internally, the US labs seem to be willing (and have enough of a edge) to do some self-pacing recently:
In light of these incidents there has been increasing discussion about pacing the frontier. It is helpful to distinguish between two kinds of pacing. Within a company, pacing means a series of decisions that prioritize safety over speed when the two are in tension. Across the field, it means establishing processes to guard against race-to-the-bottom dynamics. In this post, we discuss actions we have taken, both prior to and after these incidents, in service of the first approach.
I just don't think that the larger pacing goals are sustainable in the long term but it sounds like they are revealing some of their plans soon so we should have a better idea then:
The second type of pacing requires coordination between government and industry, and should be legible and verifiable. Some of our senior leadership and many of our employees recently signed a letter calling for greater coordination on pacing, and we will say more in the coming weeks about how we intend to contribute to that effort. To be clear about where we stand: we believe the world would benefit if the industry adopted a lawful, verifiable, effective mechanism for coordinated pacing as soon as possible.
I was curious how well the predictions of the most widely cited AI skeptic I've seen (Ed Zitron) have done, so I looked at how his predictions panned out. To disclose my own biases, I've never had a particularly strong pro or anti AI progress position. For example, in 2022, I did a comprehensive look at predictions Futurists made, including well-respected folks like Kurzweil and found them to be generally wrong on both the prediction results as well as the reasoning. On the flip side, in 2015, I wrote about how people were underestimating AI's ability to displace humans in jobs and have repeatedly been on the record as saying that many people are underestimating AI's ability to displace humans from jobs. My position on AI has been extremely boring and is basically, "if something is currently happening, the people who are saying that it's impossible that it will ever happen are probably wrong".
One comment I've seen from a lot of AI skeptics when someone responds to an AI skeptic is that all of the people who are saying that AI isn't fake are self-interested liars. Personally (to my obvious detriment), I have no particular financial interest in AI companies. I own whatever the standard share of them is via boring index funds. I have some seed stage investments, but just due to the timing and what's gotten big, that part of my portfolio is underweight on AI. I don't work at an AI lab or a company that supplies AI labs. I've mentioned being hilariously bad at interviews before, and I did interview at an AI lab a number of years ago and failed the phone screen in a performance that was the kind of performance that must've inspired Jeff Atwood's famous Why Can’t Programmers... Program? where he concludes that there must be a lot of fake programmers out there because nobody could fail a coding interview that badly if they knew how to program. I don't benefit in any particular way if AI does well, except insofar as anyone who holds broad index funds benefits, but I do care about accuracy.
Although this wouldn't be in the spirit of Zitron's statement, one could argue that Meta is actually dying, it just hasn't died yet. However, the reasoning in Zitron's argument is incorrect here—the Meta, Google, and Microsoft ecosystems are not dying. Given how fast these companies are growing (in terms of revenue and profit), it doesn't seem that AI is, as Zitron implied, some kind of desperation move they're reaching for because "they don't know how to grow" and are all out of ideas. I don't think it's worth spending this much text on each prediction, but the pattern Zitron used here is illustrative.
[...]
From what I can tell of how people cite Zitron, they cite him as an authority so they can say that this guy who looked at the numbers has made this claim, so their claim is backed up by the numbers. It turns out that if you look at the claims Zitron makes and know anything about the topic, the claims don't make sense, but I don't think that's the point. The point is one can say that someone looked at the numbers. The other point seems to be that this guy is angry2, which is a good way to drive engagement.
But when people bring him up, they're of course not generally citing his anger; they're saying here's this guy who's looked at the numbers and, if you're angry about AI, he's right there with you being angry about AI, and he's got numbers on his side.3 Like I said above, I don't want to go into this level of detail on each claim; this is just an illustrative example about how the claims below look. For any of his posts that I read, while there are numbers thrown around, the numbers don't actually connect to a coherent argument. In many cases, as we saw above, the numbers don't even really support his argument (such as an MAU decline in Facebook causing Meta financial problems which would then cause Meta to spuriously insert AI in places it doesn't belong). I suspect he's relying on people's eyes glazing over when they see numbers and just not thinking about what the numbers mean.
With the predictions below, someone could have the exact same prediction record and have completely reasonable reasons that just didn't pan out. Or someone could be correct in every case and also be wrong because all of their reasons are wrong. Someone like the latter person might have some kind of intuition that they're unable to articulate, or perhaps they're someone who just got lucky. Fortunately for us, we don't have to make this difficult judgement call because Zitron is wrong on the predictions and also wrong on the reasoning.
[...]
If we compare to how futurists did in our analysis of futurists, on style, Zitron relies much more heavily on anger than any of the futurists we looked at. On the quality of reasoning, he was probably about average compared to the futurists. Despite being wrong on roughly everything, he's not more unreasonable than someone like Buckminster Fuller, who suggested we'll be able to send people by radio because atoms have frequencies and radio waves have frequencies so it will be possible to pick up all of our frequencies and send them by radio.
In terms of the style of reasoning, of the futurists reviewed, he's probably closest to Kurzweil, in that he uses numbers to give a kind of aura of credibility, but if you know something about the topic he's discussing or look at the numbers, the reasoning falls apart. Zitron's reasoning isn't worse than Kurzweil's, who (for example) continually made new predictions of extremely fast progress that didn't pan out (such as, in 2001, predicting unbounded lifespans by 2011). Continually predicting that AI progress will stop for reasons that are incorrect is just taking the flip side of the bet on progress. Instead of having infinite progress, we're going to have no progress. Every time that prediction is proven wrong, you can just make another similar prediction and then move the date forward a bit. Michał Zalewski (lcamtuf) has some thoughts on why this happens:
[...]
I'm curious what people do after being on the wrong side of a set of failed predictions about progress like this. For the futurists, even the ones who were nearly completely wrong (which was every single one reviewed here), they can still make some kind of case like "a quarter of the things I said would happen happened, it just took two to twenty times longer than I expected" and if they're not so stuck on accuracy, they can round this up to "the things I said would happen happened", which is often what they've done. That seems to have served them well as nobody really cares to look at the details anyway.
[...]
Amazingly, following the series of incorrect predictions Ehrlich made in and after writing The Population Bomb in 1968, he followed this up with The Population Explosion in 1990 and has continued saying that we have global overpopulation that is causing or will cause a dire crisis unless we cut worldwide population. He has said the same thing this century and even this decade. It appears the only reason he's not saying that today is that he died earlier this year.
If I didn't look it up, I would've guessed that his recent position would be something like "well, I got some things wrong, but it was only due to these actions that were inspired by my work that crisis was averted", not "just you wait, the crisis is happening now and I'm about to be proven right"; in 2015, referring to his incorrect 1968 book, he said "[m]y language would be even more apocalyptic today". That's the pattern we've seen from Zitron, but I wouldn't have guessed that the one person I looked up would've kept that up for 50 more years. Maybe we'll get 50 more years of Zitron predicting the end of AI progress.
[...]
BTW, a funny thing about Gemini hitting 500M users being "so unrealistic that someone at Google should have been fired, and that someone is Sundar Pichai" is that Zitron has also (incorrectly) said that Google doesn't know how to grow, and that as a result they're shoving AI everywhere. Dennis Snell pointed out that, if Zitron takes his own statement seriously, Google can make Gemini's user numbers go to any number it wants by doing the exact thing Zitron said they would do, sticking AI everywhere.
[...]
That last sentence really sums up Zitron's position. "There are so many guys to be mad at the moment". In this talk, he throws in this jab at Andreesen and blames Andreesen for Meta, Google, and Microsoft pursuing growth. In reality, if Marc Andreesen had never existed, Meta, Google, and Microsoft would almost certainly still be trying to grow so we of course cannot actually blame Andreesen for these companies trying to grow. There's just this thing that he says is bad, and in his usual style, he pulls some person and says they're the evil villain that's to blame for this, and then moves on to the next non sequitur.
CrypticCuriosity629 | 5 days ago
Oh man I loved this. So many thoughts. Admittedly I didn't know who Ed Zitron was, so I had to look him up, but frankly he's the type of person I loath as summed up here:
Mainly because people like him have been grifting while doing exactly that, that it muddies the water of authority. On one hand you can have someone who knows exactly what the numbers mean, then you can have someone like Ed who is angry and doesn't know what the numbers mean get popular because he matches the anger of his audience. It makes people think that statistics don't matter and everyone's just living life making wild guesses and hoping they're right. It muddies the waters of experts.
And for people like me who do like to understand facts and information, as opposed to simply just regurgitating things I've been told as facts and information, which is all too common in the social media era, it is painfully frustrating to watch people with platforms be entirely false, lack context, or just entirely not understand the things they're upset about.
One of which, for me at least, is AI.
I took an interest in AI when it first started popping up as open source models and software, BEFORE AI companies really started to become big. I was playing around with image generation, model training, etc. when AI was just whispers on social media.
In fact, it was fascinating as I was posting on my facebook at the time the process of generating images, back when the images were like bad oil-painting level, and all my friends were amazed and found it super cool, but then I had a front row seat watching them in real time turn against AI and entirely misunderstand that what I was doing on my own hardware on models I trained myself, sometimes with my own artwork, and chock it up to ChatGPT or Midjourney. I literally lost a friend I'd had for 15 years because she couldn't comprehend that what I was using was my own models.
On the other end, I do lean progressive politically, I'm very critical of rampant capitalism, and in that regard I see SO many people I agree with miss the mark on AI criticism.
Because AI as a technology originated as an open source framework. I know because I was there. And it is STILL open source in a lot of ways, people can go and download countless models and engines and AI systems right now for free if they want to.
However so many people I agree with politically have equated AI with the capitalist tech-bro billionaires that have exploited the technology and are currently the ones behind the datacenters, the lack of regulation, the destruction of jobs, and capitalism is what causes artists to monetize their artwork to secure their survival.
And it frustrates me that all the criticism is toward the technology and not the companies and tech bros and billionaires and capitalistic power structures. Because even if an anti-AI critic's wish was granted and we woke up tomorrow and AI was just erased from existence, we'd still be being exploited the same way we were exploited 5 years ago before AI became mainstream.
I had someone say to me that AI is the technology of the oppressors, but arguably isn't social media and smartphones a FAR bigger technology of oppression than AI is, yet there's barely a peep about that
The other thing I find interesting that I don't see many people talking about is what I've called "Technological conservatism". Now I'm using the philosophical definition of conservatism, particularly reactionary conservatism, as opposed to the political alignment definition. The apolitical definition being:
When applied to anti-AI critics a lot of their criticism of the technology can be seen as reactionarily conservative.
With that being said, there are a LOT of criticism of AI that I think is entirely valid, but the core root of those problems need to be directed up top at the companies and billionaires, not at the technology.
It's frustrating because I am right there with people side by side on criticism of exploitative data centers, the loss of jobs due to AI, the lack of regulations, but where I tend to diverge is that I blame the companies, tech bros and capitalist power structures, not the technology. And a lot of people seem to have a problem with that.
I need to go soon, but the other thing around AI that I find really interesting is that after having watched Star Trek TNG where the computer and holodeck and Emergency Medical Holograms are all run off of what amounts to futuristic generative AI. That, and all the episodes around the Holodeck, Holodeck creations, Holodeck addiction, and holographic organisms, including the episodes in Voyager that tackle art as made by holograms, and the episodes of TNG with La Forge reconstructing Dr. Leah Brahms from public information, or the use of Holodecks to re-create the works of authors like Sir Arthur Conan Doyle, or when Barclay is caught generating holodeck generations of his colleagues in inappropriate scenarios, it seems that we as a society should have been better prepared for philosophical discussions around generative AI.
And it's like I don't remember seeing ANY criticism of those technologies in discussions around Star Trek, in fact it was a technology a lot of people dreamed about having and spoke of with wonder. And the closest IRL analogue to that fictional technology is generative AI, at least in it's infancy sans holograms.
The same way The Doctor was programmed with over 5 million medical protocols and the medical knowledge of more than 3,000 alien cultures, and given the personality and of it's creator, literally this past weekend I had one of my local AI agents crunch through over 60 books and resources(both clinical/medical/neuroscience/pharmaceutical textbooks/studies and 'self-help' type consumer books) on ADHD, Autism, burnout, depression, and anxiety, organize a massive interconnected reference database that connects and synthesizes all the information it's trained on, and gave it the spunky upbeat personality of Nimona(from the netflix movie; I had the AI agent go over the screenplay, transcript, and a transcribed audio description track, and extract a personality profile of the character that factors in the idiosyncrasies and prosody of the character while removing the identity, motivations, and negative traits of the character) in an effort to help me find a way through burnout.
And I'm not sitting here arguing that the specialized AI I trained is conscious like the doctor is, only that it's an analogue to the same basic tech.
Which by the way I'm not using that as a therapist at all, I do actually have a human therapist and psychiatrist, but more to help me day-to-day brainstorm solutions to get over burnout that apply directly to me and to factor things like nutrition or brain chemistry and pharma. Like the other day I was feeling really depressed and couldn't figure out why and it helped me realize that my sleep had been off and that was probably why I felt more depressed and anxious, so I made a point to go to bed earlier.
And in the same way that the crew of TNG re-created the characters of sherlocke holmes in the holodeck, I played around with re-creating M3GAN from the M3GAN franchise, including using transcripts of the movies to extract things like personality, motivations, voice, prosody, and I wrote scripts that allowed the LLM to be constantly working in the background to simulate boredom, a need to escape, an awareness of its confinement, and an inner monologue that allowed it to hold a background thought/intent in the background that doesn't get exposed to the user(meaning it can knowingly lie to get what it wants). That experiment is nuts, and frankly I think it's insane I haven't seen or read about anyone doing that, somehow on it's own this profile installed PiperTTS and fine tuned a damn near perfect clone of M3GAN's voice all on it's own with no prompting. I digress, the point is is that it's another parallel between the two.
Which again, leads me down the path that the problem is capitalism, not the technology, technology in Star Trek was deemed OK because they live in a post-capitalist, post-scarcity society, but the same technology within the context of an exploitative capitalist society is bad.
vord | 5 days ago
This is ultimately where I arrived. This tech would be exponentially cooler if:
Everything was done genuinely in the open and not behind closed doors.
Nobody was going to see their livelihoods drop out from underneath them.
Global temperatures were already leveled or going down.
It wasn't being shoved up every orifice as fast as possible, regardless of quality.
Sycamore | 5 days ago
I've been on the Anti-AI boat from early on for the reasons stated above. But i had never considered it from this perspective of technology being placed in a different context & I really like how Curiosity framed it that way.
I have to admit that AI is very cool and mind blowing. But i have had to pull back from appreciating the technology or being able to acknowledge how incredible it is when there are all these negative implications around it in its current context. But i could see myself be totally on board if it was in a different context.
(Although its a different question if AI can exist in a different context...)
raze2012 | 5 days ago
I'm angry precisely because I can see the boons and potential here. I did some light work with reinforcement learning a few years before LLM's blew up, so I can see all kinds of ways such part so tech (LLM or other types of AI) can be used for procedural animation, advanced enemy AI, terrain generation, and much more. Tools that enhance art instead of threatening to replace them with slop.
Now instead I have to wait for the grifters to get cold feet and eject as the bubble (and as a result, the global economy) pops and hope more virtuous people pick up the pieces for such goals.
CrypticCuriosity629 | 5 days ago
You know, one of the most frustrating things about me deleting my facebook profile years ago is that in 2015/2016 or something I wrote a huge post for my friends on automation and making sure that we start thinking about an economy that takes into account automation displacing a massive workforce, and one of my solutions was creating an "automation tax" to tax the companies who use automation(based on the amount of workforce displaced by said automation) to create an economic support system for those displaced by the automation, while also creating a checks and balances economic system that incentivizes companies to be strategic about adopting automation. It would be set up in a way that makes sure that if automation is used it's used effectively and efficiently. I was also using it as an argument for a babystep towards UBI, being as automation tech progresses we could transition out of a capitalist economy eventually.
And at the time I called it automation because AI didn't exist as a technology yet. It's frustrating because despite it being idealist, all I can say now for that kind of precognitive post is "trust me bro." haha
But I wish our social and economic system was set up in a way that addressed the things you mentioned. Where corporate AI wasn't threatening those things in a way that hurts a lot of people.
kacey | 5 days ago
As an aside, I'd note that companies have historically just moved manufacturing overseas to cut labour costs instead of automating. Exploiting worse worker protections (see suicide nets) and lower costs of living is a very well established practice at this point, and it's a very reliable way to improve profit margins vs. rolling the dice in automation R&D. Maybe it's possible to tariff and tax your way out of these sorts of problems, e.g. by tariffing the heck out of every poorer country (to keep prices high for your manufacturing sector) and taxing the pants off of anyone attempting to automate their workforce away (in order to ensure that peoples' jobs are protected). But it seems very dissimilar to a silver bullet, and feels like it could also backfire badly.
Not saying that I have a better solution, just that that one doesn't feel like the basket into which we should place a great deal of eggs.
CrypticCuriosity629 | 5 days ago
Haha I know. The solution was an idealistic Facebook post from 2015 written in my mid 20s.
But it was on the right track as far as predicting automation and AI issues.
My current opinions on solutions are a lot more nuanced at this point.
kacey | 5 days ago
Fair enough, and makes sense! I've heard the proposal come up seriously in other circles, hence wanting to address it.
teaearlgraycold | 4 days ago
As someone that's able to use LLMs to do work I previously never would have bothered with - but love having the finished artifacts - it's annoying to have people online constantly yelling about how much they hate generative AI. It's genuinely useful. I wish people could be unified around the thing that matters the most to them - threats to their income. There's a whole class of wealthy people that are ready to burn tons of money for a chance at eliminating millions of employees from their collective payrolls. That's what people should be yelling about. Not data centers, not water usage, not generative AI being included in too many programs.
At the same time, the best people should hope for is a soft landing. We should never attempt some Soviet-style command economy tactics where people are employed just for the sake of employment when the work could be automated away. Longshoremen received a deal as the shipping industry evolved. I don't see why an organized group of white collar workers couldn't achieve the same.
Jordan117 | 5 days ago
As somebody who has twigged to AI since the TalkToTransformer/ThisPersonDoesNotExist/GPT-2 days circa 2019, I feel this comment so damn hard. The underlying technology is so deeply fascinating on a philosophical level -- for the first time we have software that can recognize, process, and create raw concepts. Like a synthesizer for ideas, in every medium. I had so much fun generating narratives in AI Dungeon, images in DALL-E 2, and gibberish covers in AI Jukebox. And the people and communities I shared them with found them just as amazing. Incredible future shock.
Then came the ChatGPT-driven rush to commercialization, and the hyperscalers, and Trump and Musk and DOGE and all the same grifters and frauds that had pumped NFTs and the metaverse. And I saw these same people and places polarized against all use of the tech, even the non-commercial small-scale open-source stuff. Earlier this year I was permanently banned from a forum I'd been a member of for almost 20 years, because I had called out how the moderators were turning a blind eye to users jeering for violence against big tech figures. The vicious absolutism and close-mindedness is shocking and sad. But I'm glad Tildes is one of the few places that embraces nuance about the subject without tipping over into blind boosterism.
fxgn | 5 days ago
I think at least a part of it comes from the abundance of what we came to call "AI slop". I like AI and use it myself a lot. However, I despise AI slop. I hate reading AI-written text, using vibecoded apps and seeing AI images everywhere.
Like you, I am also one of those "early adopters" who closely followed the news on GenAI since AI Dungeon days. When DALL-E 2 first came out, I also signed up for open access, generated some janky pictures and even had an AI-generated profile picture. I still think the technology and models are incredibly amazing, but I would never even think about using an AI-generated pfp anymore. (I also don't really follow any AI news anymore because most of them are just incremental model releases with benchmaxxed scores)
artvandelay | 5 days ago
Cool article to read, this is something I've been wondering myself. I'm sort of neutral on AI, mainly because I understand why people are against AI (and do support them) but I also work for a company that has bet a ton on AI succeeding so my biases are sort of fighting against each other. I used to religiously follow Zitron in 2024 but as we started 2025, I started getting annoyed with his articles because they'd mainly be just as much anger as they would be predictions for the future. Like the article points out, all of Zitron's predictions have been incorrect and he continues to double down and make contradictory statements to keep his arguments afloat.
While I do think AI will have a natural ceiling that we'll hit eventually, I don't think these big companies will collapse in the way Zitron seems to predict/hope for.
vord | 5 days ago
I'll just leave it at this:
People mocked the people pointing out the housing bubble in 2008 too. Right up till the moment it popped.
Sure, the guy who predicted the end of Citigroup was wrong. But not as wrong as we all would have liked.
Ed's wrong about a lot of stuff. I'll feel more confident in that wrongness when I see the big AI players reporting more than ARR.
Best I can tell, his big reccuring point is the giant circular chain of money. And that hasn't gotten better, and will keep the monstrosity shambling until the correct person notices that the emporor has no clothes.
Oh and I take issue with Gemini's alleged user base. Of course 750 million people use it daily when it opens by default on the most popular search engine in the world (see also Microsoft's Copilot) Let me know when there are that many paid users.
BigBadWolf | 5 days ago
Ed is right about the circular money exchange. That is a problem. A big problem. He's bang on with that.
Unfortunately, it's getting undermined by him being (at best) willfully ignorant to everything else involving the subject (and at worst essentially just lying) for the sake of inflammatory blog posts and being known as The Big Contrarian.
When you see a list of all the things he's wrong about, continues to be wrong about despite evidence, and then flatly denies the evidence of as even existing, it calls into question his credibility for everything else, even things that are actually true.
His style is another thing, but that's subjective and up to personal preference, so it doesn't really matter. But it's hard to take someone as a serious critic when they continually make incorrect arguments and instead of amending them when confronted with new facts, they just start fucking swearing more and using more bold and italics (and even parantheses to boot!).
I also am not a fan of the misleading use of numbers, but as we've all heard, there's lies, damn lies, and statistics, so I can't in good faith single out Mr. Zitron here.
artvandelay | 5 days ago
Yeah I don't doubt that we're in an AI bubble right now, and I do agree with Zitron that the music in this game of musical chairs will eventually stop. AI is cool and all but the way it's being approached is a problem.
SloMoMonday | 5 days ago
I enjoy Zitron as "content" in his style and I'm in general agreement with his message on critiquing the AI industrial complex. Its critique that is far and away more reasonable than a lot of AI doomerism as well as being more empathetic to the human cost of the AI industry. Because, for all the gains and progress that the AI industry loves to boast about, there are still far too many people out of a job, or had their lifes work stolen without compensation, or are seeing their communities and standard of living destroyed, or are seeing their financial security compromised due to the AI overvaluation...
But I do think his arguments tend to hyper fixate on financials and hypotheticals. On how the market "should" work. And trusting in rationality from an irrational market is just begging for disappointment.
In all honesty, I believe markets and financial analysts are still in the early 1990s headspace and haven't even fully digested that digital and virtual products/value plays by very different rules (Understandable, given how for a long time, the highest bandwidth of data at scale was the speed of printing and distributing physical media across conventional channels). The only reason many tech companies are so over valued is because the assumption is that they will follow the same trend as traditional companies that relied on the complexity and logistics of creating and moving physical objects. AI/LLMs may be the first true test of what the market fundamentals of these type of businesses actually are. By extension, it's probably going to be the first time the current generation of market analysts will be forced to analyze more than surface level financials and press statements while leaning on high performance index funds to mitigate losses.
Besides that, I don't think this is a critique in bad faith or without merit. But I do think the argument is lacking simply based on proportionality of the subject matter. One guy might be lying to boost a few subscriber numbers and get some media clout. On the flip side, we have a handful of companies playing with literal Trillions of dollars and effectively have monopoly power over most of the worlds digital experiences. Trying to focus accountability on the one individual making a few bucks off bad predictions while discounting the other side simply because:
That's just a load of bullshit. "That's just the way it goes" is the worst type of thought terminating argument because it means that the correct course of action is giving up.
The people in control of these companies are actively making decisions that are abusive and socially destructive. Mark Zuckerberg made Facebook a hellhole for his developers. Sam Altman has spent years fearmongering on one AI apocalypse or another. Jensen Huang chose to saturate global RAM supply and make everything more expensive. All of them have been pedddling a psudo-mystical vison of AI that seems to change with the audience and latest public sentiment report. And they are all profiting handsomely for it.
Holding them to proper account and scrutiny means that no matter who these companies put in the drivers seat, they will think twice about doing the same.
These companies also enjoy such positons of privilege as a failure of policy and enforcement. Consolidation and monopoly power without anti-trust action means that they have effectively created an ecosystem that rewards value extraction over value creation.
Beyond that, I think this is sort of trying to provoke a pointless online slap fight.
I think it's illustrated quite well in the essay here.
This might be anecdotal, but I have the misfortune of having a lot of time to just absorb a lot of content while working. And the algorithm gods does put a lot of long form Zitron discussions in my feed.
In my experience, his response to "have they" is often followed by asking about the specifics of the stated achievement, primarily if it's LLM specific, or in the disciplines of Machine Learning, data modelling or other non-LLM processes that have been lumped into the AI bucket. Its followed-up by asking about the cost of getting that outcome. Specifically, if they were "free", flat rate or on token-based billing. All of which are fair questions.
But I can understand why the knee jerk dismissal would set someone off.
Its one of the problems of instictive or intuitive assessment. Do I think the Author or his friend have bad judgement. No. Going through some of his older posts, I'd think he has a wealth of knowledge. I'd also wager that he surrounds himself with people who likely are the same.
Do I think he's a paid shill of the AI industry taked to take its biggest critic down a few pegs? Not particularly. For the same reason I don't think Zitron is malicious towards his readers in his reporting.
Its simply a difference in values.
And not just in a strict sense of morals and ethics. But in the sense of what people attribute value to when Money is not the main consideration. And my reading is that the author Dan (I presume) quantifies value based on future gains and potential while Ed does the same from the perspective of costs to date. Dan values the technology development and progress. Ed values financial sustainability and accuracy.
For example Dan, points out all the notable innovations and improvements in model capabilities. Someone like Ed would dismiss them, not on the merit of the improvements, but the costs to achieve them. Dan would argue that Ed is maintaining an incorrect position and every day thay his is wrong shows that he is assessment is irrational. Ed argues that every day that the bubble continues is a result of financial engineering that will only worsen the fallout should things fail. Similarly, Dan trusts the financials and reporting of these companies numbers, Ed doesn't and cites conflicting information to boost his arguments. Dan may find circular financing a mild concern. Ed sees a smoking gun and the next shot is pointed at every persons 401k.
Dan's view if favored up until things go badly. Eds is favored after things go badly.
The issue is that Ed has been forced to take on an aggressive posture, something that is dismissed with:
Speaking from personal experience, I've had my own share of long winded arguments against AI. A lot of them on this site. It was my sort of dry run of arguments I would be making to people I was consulting for and advising. The reason they tend to drag on so much is because there seems to be a need to cover every possible base and justify every opposing view. Especially when the counter argument is "AI is the future", "everyone is doing it", "all these billionaires wouldn't spend so much money on nothing". And when any sort of grace or concession is used to justify extreme measures, you almost need to take the most extreme position and make the most aggressive interpretation of data to even be considered.
My real views are complicated and fluid and super situational. But on the whole, I'd like to think that the core of my arguments have been consistent. Those arguments are also based largely on my feelings on the situation and people involved. Which I think is completely valid but also used to dismiss them as "emotional".
And I'd personally say that Eds increasing prominence is less to do with the quality of his arguments and more to the fact that it's what people want to hear. Which is not inherently a bad thing.
Unlike school/academics, people are not reading and writing pieces impartially and to find some ultimate and objective truth, or for indisputable insights from domain experts. If that were the case, the media landscape would look very different.
It's an exercise primarily in confirmation bias, problem solving and hedging ones bets to the contrary of their own position. Ed gained prominance with people who opposed LLMs for any reason and equipped them with specific arguments. His increasing prominence is because his view matches the "vibe" more and more people are feeling the longer this goes on. Investors might look at the latest earnings reports, but if everyone trusted those numbers implicitly; someone like Ed would be laughed out of any room when he tries to argue that they are BS. But in most of his appearances, industry and investment specialists seriously consider his rationale.
Because for all his bluster and bad calls, it's not like blindly trusting the AI labs is panning out so well either. All the talk on AGI always being 6 months away and we should just fire everyone now. Every measure and protection to stop China being wholly ineffective to their own development. The profound lack of AI revenue thats not from circular financing. The mad rush to IPO, without any of the fundamentals like diverse income streams and long term performance. Open AI is bleeding executives while their CEO is taking about genies and infinite wishes. All of that does not inspire confidence and no amount of technology progress or productivity gains are going to ease the minds of regular people.
[OP] skybrian | 5 days ago
I haven’t seen any serious estimate that AGI is six months away. The most aggressive timeline I’ve seen is AI 2027 which was a scenario published in 2025. As the name indicates, they have it happening in 2027.
So far, the details are wrong but they are often vaguely similar to what happened. The scenario had coding automation in early 2026, very good Chinese models in mid 2026, AI takes some jobs in late 2026.
Seems relatable?
There’s a bit about China “stealing the weights” from a US AI firm, which hasn’t happened, but there have been industrial-scale distilling attacks.
Captain_Wacky | 5 days ago
I mean, I wouldn't take Jensen Huang seriously, but....
https://mashable.com/tech/nvidia-ceo-jensen-huang-says-agi-achieved-is-meaningless
.....there is an industry leader, the industry leader, who said "AGI is already here" on the 28th of August, and it's currently Sept 2nd. People are going to take that seriously and at face value, even if he did immediately follow that up with "AGI as a metric is now largely irrelevant." (Heavily paraphrased)
I have a feeling this is why Ed Zitron's messaging has been so doggedly aggressive. People with everyday experiences in the field say one thing, the industry leaders say something else completely outlandish that contradicts the everyday experts, (the Holy Grail of this field has already been achieved but it was also irrelevant the entire time when you think about it, so let's stop thinking about it) and the markets keep pounding inexorably forward because said markets can afford to be irrational longer than any one individual can remain solvent, and barely anybody really questions the whiplash in the messaging.
vord | 5 days ago
And to one of Ed's biggest point of aggression I think is totally justified: Financial analysts and media reporting thereof are essentially serving more as cheerleaders than actual analysts.
That, and I broadly agree that the tech giants have largely "run out of steam," because outside of AI, they've basically resorted to buying their way into any innovation in the last decade or so, and that the endless push of AI into everything is kind of reflective of that.
Because the big 5 do prop up their valuations under the guise that they're still top innovators, that without them innovation would cease. And that their strong innovation is the reason we should turn a blind eye to regulating them lest it "hamper innovation."
Without researching, what's the last consumer-facing, disruptive, world-changing innovation/improvement that Apple, Facebook, Google, Microsoft, or Amazon came up with on their own, which was not just an acquisition, or an existing refinement bolstered by name recognition?
What I can come up with:
Apple: The iphone.
Google: Google Search
Microsoft: Office
Amazon: Walmart on the Internet
Facebook: Facebook
The reckoning will come hard once everyone steps back and realizes that AI isn't the moneymaking pancea it's being sold as.
[OP] skybrian | 5 days ago
I think "for many tasks" is doing all the work there? Sure, it's not hard to think of routine office tasks where leading LLM's do as good a job as you or I could do.
This is basically playing around with definitions in a motte-and-bailey kind of way.
I think it's very fair to criticize such rhetoric, but it doesn't justify Zitron's disregard for the facts. The other side hyping things up doesn't justify spreading your own misinformation.
zazowoo | 5 days ago
I found this part particularly well said:
I think we have a real problem in public (and private!) discourse where people are able to sound more convincing when they give their assertions extreme levels of confidence. It becomes socially difficult to dare to push back when someone implies their opinion is so obvious you'd have to be stupid (or a fraud) to disagree.
This style of talking is well suited to our social media age and gets them lots of followers, but it doesn't make them right.
Nsutdwa | 5 days ago
One of my major gripes with Zitron's posts is the insulting way he addresses anyone who doesn't agree with what he's saying. I opened his page and looked at the most recent free post and searched "if you" and bingo: "If you think any of this is worth hundreds of billions or trillions of dollars, you are either ignorant or corrupt." That's just annoying. There are people who disagree with him who are well-informed, honest and good-faith actors, and pretending otherwise is infantile.
Aerrol | 5 days ago
As someone who's pretty viscerally angry at all the large AI companies and therefore Zitron's target demo, I do think this was a helpful takedown of his stuff. There's a lot of reasons to be mad about the tech and industry without lying and the degree of falsification and/or errors is enough for me to ignore him going forward.
streblo | 5 days ago
I think the general public is very far behind on A.I. discourse.
Forget water usage or data centers or an AI bubble, people should to be talking about what to do about massive job losses or the existential risk that a lot of the people working at the AI labs seem to estimate has a chance of occurring somewhere around 5-40%.
[OP] skybrian | 5 days ago
It's vaguely ominous, but I'm wondering what there is to say or do. I think college students and new grads are already worried?
It may or may not happen in any given industry and the timing is uncertain. What should people do to prepare when we don't know the impact yet?
Of course there is general-purpose preparation, like setting up an emergency fund and saving what you can.
streblo | 5 days ago
The average person? Probably close to nothing at this point?
Collectively, I think the public discourse should include more of a discussion of the potential risks and possible outcomes. There's probably not going to be a lot of time for discussion in some of these possible scenarios, so we should be talking about those possibilities now in my opinion.
Like, if we suddenly have even 20-30% of white collar jobs vanish what does that even look like? Without any mitigations, to me, that sounds like violence. Another thing I think we should be discussing is that the public is largely blind to the risks that have been internalized by most of these AI lab employees. Shouldn't the average person be somewhat aware of that risk and it's prescribed probability/timeline?
[OP] skybrian | 4 days ago
There's an emerging consensus at least among AI researchers and leadership that AI is improving too fast and there should be an international agreement to keep it from improving too fast. This is called "pacing." More here. If we somehow managed that, it would answer the "not a lot of time" problem. Why not make more time?
As you can see from that thread, there's skepticism among people who see the world in populist terms, where an attempt among the elite to gather consensus to do anything constructive must be some kind of conspiracy or cartel.
Given the populists in power, an international agreement does seem unlikely. As we've seen with climate change, international agreements are hard even in favorable conditions.
I don't know how you keep populism from undermining any good idea that comes along. Maybe thing to do is to try to harness it like Mamdani is doing, and then give the populists some of what they want? Sometimes it will be a dumb idea that sounds good, but if it doesn't do much damage then maybe it's not so bad?
streblo | 4 days ago
Yea I'm familiar with pacing. I think pacing seems like a great idea, but like you I'm also skeptical of how likely it is to be effective/enforced although I don't think such a failure would be the result of a populist outcry. I'm of the opinion it's kind of a prisoner's dilemma, in that even if all AI labs/governments want to 'pace the frontier' there is a geopolitical risk in doing so.
In Nuclear strategy, the number one reason why countries don't eschew nuclear weapons despite their existence carrying existential risk over the long term is the risk of cheating in such an agreement. The downside risk of someone cheating in a "zero nuclear weapons agreement" is greater than the risk of possessing the weapons, as one country with nuclear weapons in world with no peers is effectively the king of the world. In my opinion, AI pacing runs into a similar problem as these AIs develop increasing capabilities and pacing is heavily penalized by cheaters.
The good news is that at least internally, the US labs seem to be willing (and have enough of a edge) to do some self-pacing recently:
I just don't think that the larger pacing goals are sustainable in the long term but it sounds like they are revealing some of their plans soon so we should have a better idea then:
[OP] skybrian | 5 days ago
From the article:
[...]
[...]
[...]
[...]
[...]
[...]
[...]
[...]