I've said this before in here but Zitron is completely captured by his audience at best and a grifter at worst. A couple of months ago he was tweeting that people were crazy talking about _agents_, that they didn't _exist_ and that people talking about them were shills or bots. The responses were full of incredulous software developers saying but but but I use them every day, they're so good they're scary actually...

Zitron is just, like Dan Luu says, wrong about everything and doesn't care anymore, he's in the business of extracting money from his engagement.

Accurate or not, there are many risks other than those that could bring AI companies value to 0. E.g., systems not requiring passing over all the data continuously, but in small regions running at L2-L3 cache memory speed. Once someone figures out that, the need for high bandwidth memory would end.

I don’t know much about Zitron, but YouTube somehow pushes these “experts” onto my feed. This guy is full of crap, he doesn’t even care, because that doesn’t matter, what matters is view count.

How has it taken so long for this to get some proper attention?

I'm not Ed's biggest fan, but he's been pointing out some very important things for quite some time now, regarding the ludicrous over-investments going on around so many companies that are completely and utterly devoid of profits and will likely never have any.

The author asks "How can people take this seriously?"

I would reply the same regarding this article. Nearly all of these refutations are unconvincing.

The claim: "I believe we're reaching the upper limits about what generative AI can do and how accurate its outputs can be."

The rebuttal: "Zitron's argument at the time was that hallucinations were as good as they were going to get, which meant that AI performance is capped at 2024 levels. Both the overall prediction and the mechanism were wrong. This one seemed wrong at the time, in that I noted here in 2024 that you can make AI code halfway decently by just putting it in a loop and having it run until the code compiles and tests pass"

Putting an LLM in a loop, burning tokens, and thrashing against a compiler and test suite is a ridiculous way to say that hallucinations have been "solved". Please. This is absurd.

I've always been suprised with the amount people take him seriously. I think he's someone who makes people who dislike AI "feel good". I don't blame someone who doesn't understand AI or have read what he's written.

But for the bloombergs and other podcasts I would have expected them to do a bit of research. I honestly think with the amount of doubling down he's been doing that he's a grifter.

Obviously Ed is a special case, but let's be honest, pretty much anybody telling you they can predict the future is selling you BS. And that includes all the people confidently predicting he was wrong. The actual situation is, there are genuine unknowns driving things with significant influence, and nobody actually knows what is going to happen. At best, people can talk about risks and likelihoods of outcomes.

Mo Bitar over on YouTube does the grumpy AI skeptic routine in a much more entertaining fashion.

The simplest argument against AI is the fact no public company appears to be making money on it, minus revenue that contributes to AI infrastructure.

Would love to see broken down counter example of public company.

So far Chegg and Duolingo have been devastated. Surely they could cut costs drastically with AI?