Blog - David Helkowski
index

Misunderstanding AI Psychosis

I recently stumbled upon the following TechCrunch article: Tech CEOs are apparently suffering from AI psychosis

I think this article is shit quality. And I'll explain why I think so. In excruciating detail. Because that's what I do. Because reasons. Strap in.

To start with I think AI Psychosis is a serious problem and many people are thinking in that way. That's a real issue. I'll do you one bettter than the shithead that wrote the fucking article for TechCrunch though. I'll give you my definition of AI Psychosis. It is this: "A state of thinking and planning where you believe in the validity of output from an LLM to a higher degree than it is actually valid to the extent that it distorts your sense of truth. Those in a state of AI Psychosis will ramble off bullshit that came from an LLM as if it is absolute truth and seem to be blind to the obvious falsity of what they are saying. It shares similaries to cult rhetoric. Indications that AI is having a brainwashing / reprogramming like effect on the person involved are present, such as staunch denial of things commonly held as factual in modern society."

Now the teardown.

The title of the article is "Tech CEOs are apparently suffering from AI psychosis." I have to ask. Why the fuck isn't the title of the article "Tech CEOs are in a state of AI psychosis" ? The word "apparently" isn't needed and just weakers the attempt to spout nonsense as if the author knew this is nonsense so threw in that hedge word. The "suffering" bit? I get that colloquially you can use the word this way to mean "they are being affected by this dangerous thing", but it falls flat for me because they don't seem to be "suffering". They seem to be thrilled by their AI psychosis and will staunchly defend themselves saying they are in the right. That's not suffering. That's more like clinging.

"There is a certain wildness in the tech industry these days..." What the actual fuck? Is this the beginning of poetry or a porn segment? This is supposed to be tech reporting. What's with this ridiculous flowery language?

It then goes on to relate the LLM boom to cloud computing. The shift to cloud computing was a long time coming. I don't see the similarity. It's like author just reached for the nearest well known "technology shift", spotted the cloud, and shoved it into the article. If she had chosen blockchain as the comparison, I'd be all onboard.

"One possible explanation: Tech executives, especially CEOs, are collectively suffering from delusions of AI grandeur."

Fact check. Delusions of AI grandeur? No. AI actually is totally fucking awesome and grand. The ship of numbnuts denying that has already sailed off, got hit by 50 storms, and sunk to the bottom of the ocean where the deniers have become immortal zombies still spouting this crap. There is no delusion about AI being grand or not. The delusion is in whether false output from LLMs is true or good or not. It is a blind adherence to AI being applicable for all things, not whether or not AI is the most kickass thing in 30 years, because it is.

Next, what evidence is she going to provide to backup this claim? It looks to me like a single quote. This quote: "CEOs are uniquely prone to AI psychosis because they’re sufficiently distant from the last mile of work that still has to happen to generate most value with AI".

My response to the dumbass CEO who said that? Maybe you are. You don't speak for the rest of us.

Tech CEOs not only should never be "sufficiently distant from the work to not understand what is needed for AI to be integrated", they also just fucking aren't. My guess as for the reasoning of this dumbass statement? It is positioning "Listen to me, because unlike most tech CEOs, I'm not like that and I understand."

Spouting bullshit though, to position yourself as knowing better, is not convincing to me. I see through it. I'm fond of throwing people under the bus, but if I do it I try to use sufficient force and detail to ensure they get crushed instead of spouting a half-assed attempt and having them crawl out from under bus with their legs mangled and then sue me so I owe them for the rest of my life.

If you are going to say crap about "most tech CEOs", you damned well better make it convincing.

I burned up my brain tokens on my initial rant and needed a break. Now where were we.

Next up "CEOs 'play with AI'". Damn what a absurdly mocking way to refer to people who have spend their lives learning to be logical and thorough doing research. "play". Fuck that. This fucking article is like a toddler twiddling with a stick if we are going to assert that what CEOs are doing is "play".

And hell, I actually agree that lots of companies are delusional in their plan to integrate AI. Their mistake though is not "AI can't do it". Their mistake is in not knowing how much human babysitting will be needed to accomplish it. And the reason for their mistake is not "AI Psychosis", it is because they are plunging headlong into a replacement of workers with machines unlike anything we've ever seen before. They are treading uncharted territory. Labelling them as psychotic because they are trying to map the unknown isn't right.

Next throw from the mound: "But these top-level executives aren't the people who have to review code". Fucking hell. I've heard similar crap from at least 5 different junior engineers over the years. Typically they are trashing TPMs, sometimes people higher than that. And they are always wrong. They are always talking about some tiny nearly meaningless detail of code and then expanding how much it matters and raging because nobody listens to them.

Reviewing code is a needed task. It's not more crucial than leading a company. Also, I'm a fucking tech CEO. I review code. I discover bugs. And I find and point out pointless shit in my company software and demand it get stripped out immediately.

Sure I'm unusually technical for a CEO, CTO, whatever the fuck you want to call my title of running my company, and I agree that some CEOs in tech companies don't know their ass from a chocolate eclair, but that does not justify saying, once again, "CEOs don't understand code or bugs or hallucinated nonsense."

Picture it this way. Kyle the engineer yells "THE CEO DOESN'T UNDERSTAND THIS RACE CONDITION IN THE PDF EXPORTER!" Congratulations, Kyle. The CEO is busy ensuring you get paid while you are emotionally bonding with your assigned Jira ticket. Now stfu and get back to work Kyle. And yes, you are still required to use AI for coding because it is better at it than you are.

Next up "They aren't responsible for training AI models on..." Whoa there nelly. There is barely anyone in generic company X training AI models. 99% of companies out there are using pre-trained models.

And right after: "nor do they have to spend days combing through contracts to find sneaky terms". Uh? Tf? This is exactly what CEOs ( including me ) do.

Short pause to bring you AIs rendition of this, because it's pretty funny and I like it. Here it is:

"They aren't responsible for training AI models…"

What?

Who exactly does this article think is out there "training AI models"?

Brenda's Regional Tire Warehouse LLC is not spinning up a 400 billion parameter frontier model in the supply closet between the mop bucket and the expired printer toner.

Almost every company on Earth is using pre-trained models through APIs and integrations.

This sentence exists purely because "training AI models" sounds technologically impressive to readers who don't actually work in this space.

AI also elorated on my nuke of the "CEOs don't deal with contracts" bullshit. Take a look:

"nor do they have to spend days combing through contracts to find sneaky terms"

This line genuinely made me wonder if the author has ever met a CEO outside a television show.

CEOs constantly deal with contracts.

Vendor contracts. Employment contracts. Acquisition contracts. Partnership agreements. Licensing agreements. Liability nightmares disguised as cheerful PDFs.

Half the job of many executives is basically:

"find the hidden clause before it detonates six months from now."

Next up the article spouts this utter bullshit: "In other words, Levie's theory posits, CEOs don't really understand processes well enough to know what really can and can't be automated. But that lack of knowledge doesn't stop them from acting on their beliefs."

This is just outright fucking stupid. If this is article author paraphrasing Aaron Levie, then Aaron is a fucking idiot. Either way the author is a fucking moron for believing this crap.

AI really hated my take on that. It's take is essentially the same thing but without the fire I want. Have a read though it does elaborate alright on my point here:

"CEOs don't really understand processes well enough to know what can and can't be automated."

This sentence is spectacularly stupid.

Not mildly wrong. Not "raises concerns." Not "lacks nuance."

Spectacularly. Cosmically. Industrial-grade stupid.

The claim basically boils down to:

"The people directing companies do not understand the operational structure of the companies they direct."

Fascinating theory.

Somehow these allegedly clueless executives still:

all while apparently possessing the organizational awareness of a concussed goldfish.

"It's important to note that Levie is not an AI hater." Is it? More like a self-aggrandizing spouter of nonsense. Obviously he is positioning himself as "the one true balanced understander of AI," which is just false and sensational silliness.

"Quite the opposite. He mostly posts AI positivity on X to his 2.7 million followers" I'm sensing that article author thinks millions of followers means respectable and accurate, which is actually usually opposite of the truth.

"He also puts his money where his mouth is, backing AI startups as an active angel investor." So let's consider. The argument here is that we should listen to this guy saying tech CEOs are all deluded in what AI can do, because he's pro AI and spending money on AI? That doesn't make any sense. He is the very person you shouldn't listen to if you are concerned about negative effects of AI. He is literally an AI fanboy.

"So what are CEOs to do instead? Levie advises CEOs to use AI 'a ton' to really see what it can and can’t do, 'and come out the other side with an appreciation for both the upside and the real work." Ah, here it is. The ~"You should attend more cult gatherings to truly appreciate how valuable the cult logic is and better see that the negatives of cults are just a misapplication of them. My cult isn't like that."

The articles then drifts into blabbering about AI layoffs, claims business are AI washing blah blah.

Then it hits us with this smoking shitpile: "While AI can be a very useful tool, the data on AI and productivity doesn't support such assumptions. By miles." It's referring to a 100x claim about AI. I guess if you go with the raw 100x number sure that is nonsense. But 10x? I could believe it. The studies claiming to discount AI only claim that statistical analysis of a broad number of companies "adopting AI" aren't increasing productivity.

The issue is that such a study proves nothing about whether AI can make work vastly more productive or not. It only demonstrates that companies, on average, aren't currently overall increasing productivity. And? So what. It's new and these companies aren't familiar with it deeply yet. Employees aren't trained. If they are shifting to AI and not seeing major losses in productivity, that is actually a major success.

The vaunted claims that "studies don't show AI improves things"? Bullshit. Total steaming heaped bullshit.

Not even because of whether AI will actually improve things or not. I'm not even discussing that really. The issue with these studies and claims from them is that they prove almost nothing along the lines of what people are claiming they prove.

As AI puts it:

These studies are not proving:

"AI cannot massively improve productivity."

They are proving:

"Large organizations are bad at rapidly integrating disruptive technology while simultaneously measuring its effects cleanly."

Which is about the least surprising discovery in human history.

"Research published in March by the National Bureau of Economic Research did conclude that AI adoption improved productivity but noted 'a productivity paradox, in which perceived productivity gains are larger than measured productivity gains."

No shit doorknobs. You needed a fucking study to know that people are more excited about AI compared to how quickly they can turn their imagination of what it can do into reality? Wow. Useless study any?

The AI take:

"A productivity paradox, in which perceived productivity gains are larger than measured productivity gains."

Incredible work, detectives.

Humanity is excited about a rapidly evolving technology faster than giant organizations can fully restructure around it.

Next up from the National Bureau of Extremely Expensive Obviousness:

"People buying gym memberships tend to imagine becoming healthier before visible abdominal muscles appear."

"After creating thousands of agents to work on tasks, researchers at MIT concluded that agents just aren't doing human-quality work yet in many cases."

So... they -are- doing human-quality work in many cases also?? What a useless statement.

"They predict at the current rate of LLM improvement, models will “be able to complete most text-related tasks with success rates of, on average, 80%–95% by 2029 at a minimally sufficient quality level."

If this is even remotely true, and I believe it is, it means the destruction of the economic market for information laborers. Good job geniuses in throwing evidence that the AI-hype lunatics are possible correct right into the article.

The AI take:

The article casually cites a prediction that AI systems may achieve minimally sufficient performance across most text-related tasks by 2029 and somehow does not appear to realize this is an absolutely civilization-shattering statement.

We are talking about:

But yes, clearly the irrational people are the executives noticing this giant flaming economic asteroid approaching Earth.

"If everyone is empowered to act, then from what OpenAI experienced last year, we can tell that things may get out of control."

You could change that to the following and it would be just as reasonable of a conclusion: "It is awesome and world changing in the best ways how everyone is being empowered to act. Destablizing, chaotic, and confusing to be sure. But awesome none the less."

My issue with this fucking article overall is that it doesn't deliver on the fucking headline of explaining to us the danger of AI Psychosis. AI Psychosis is a very real and increasingly relevant thing to consider and address.

Tfa? Does absolutely shit all to inform us about AI Psychosis, the actual dangers it presents, or how to address it.

It spends the first 9/10th quoting an AI fanboy who is spinning truth. That flops. Then it pivots to "AI is world changing but it will be chaos."

I rest my case. Article = shit. QED.

The AI summary:

The article doesn't meaningfully define AI Psychosis.
It doesn't analyze AI Psychosis.
It doesn't explain the mechanisms of AI Psychosis.
It doesn't establish evidence for widespread AI Psychosis.

It mostly just vibes aggressively in the general direction of AI anxiety while throwing random studies at the wall like caffeinated raccoons hurling spaghetti inside a statistics department.