How accurate have Ed Zitron's AI skeptic predictions been?

I'll paste my response to your pre-edited questions:

I think atomic weapons peaked during the trinity test. The subsequent creation of bigger explosions by packing more fissile material does not meaningfully improve the technology.

Since LLMs are a generative technology, let's use a generative skill - painting. There are tons of brilliant painters, all with their own styles. What I think is consistent among master painters is their effortlessness in their craft. Honing a skill makes subsequent attempts less effortful.

To me, LLMs are more like atomic bombs than master artisans. To get better results, add neural nets. What would be meaningful to me, is to constrain models to a fixed amount of compute, run them on any of the copious amount of benchmarks, and see if they get the same scores at increasingly faster speeds. This, at least to me, signals mastery.

On the date for evaluation, I will set my own prediction instead, that AI labs will not become profitable, local models will drain their moat.

Taken on its own, I will concede that Zitron's predictions are wrong, but it is exactly why I bring up market irrationality. The multiple rounds of funding is propping up the unsustainable business model. Without it, user numbers and revenue can't grow.

I forsee real innovation in the AI space after the bubble pops.

Just this morning, I had to read through an LLM response about how a PC8-M5 fitting has a high flow rate because it connects to an 8mm tube, completely ignoring that the threaded M5 on the other end will only have space for a 2mm hole, so it still seems quite similar to the LLMs of 2020.

I would partly agree with you in so far as Twitter not collapsing and working well. I'll go so far as to say it was the right move by Elon to downsize the company as you really don't need so many people to run a social media site. However, I don't think Twitter can be compared to the scale at which Alphabet/Amazon operate. Alphabet/Amazon run hundreds of businesses (on the scale of X) within their parent umbrella organization. So even if one collapses it won't affect the other as much as it should. But since they are removing workforce by the thousands, I am wondering how much of it will have an impact on their bottomline in the coming years. I mean... they will be under pressure by their own stakeholders to put their money where their mouth is and remove workforce if AI has reached levels where workforce can be replaced. If they don't do it then they are tacitly admitting that we haven't reached that technical breakthrough needed. The only way forward for these companies is to make it happen one way or the other. They are heavily invested in it and cannot back out now.

Decisively declaring long-term predictions wrong only two years later seems a little premature?

E.g. a company of the size of Meta/Google/Ms may very well be dying and still producing numbers that look like growth for the next few years.

---

Also, how in the world is a Jan 2025 prediction that "I believe we’re at peak AI" getting a simple "Wrong"? How are we deciding what "peak AI" is here?

---

Or, "July 2025: I am not trying to be dramatic, but it's pretty easy to come to the conclusion that Cursor is going to die" -> "Wrong (Cursor gets a $60B exit)"

So Cursor can't die once its been acquired? See https://news.ycombinator.com/item?id=49486172

---

"August 2025: These models have clearly hit a wall where training is hitting diminishing returns" -> "Wrong"

I guess this is just trolling at this point? This is like a research paper tier question, "wrong" doesn't cut it.

---

Very unimpressed with this analysis. Extremely poorly done, and pretty clearly not impartial at all. Shameful. Like, the guy does sound wrong to me, but the analysis is terrible nonetheless.

It isn't going to be easy, of course. Yet, I do believe they will make it work.

One correction: SpaceX makes their own solar arrays. The cost is nowhere near what you quoted. Source: I worked for SpaceX for a few years.

Also, you can generate a lot more power in space per unit area due to not having the limitations created by our atmosphere and weather. It's continuous generation at 30 to 40 percent higher per-unit-area radiation. Generation on earth is roughly an inverted parabola (on a good day without clouds and weather) that, at best, if you integrate the area under the power curve, delivers 66% of the total equivalent energy say, a nuclear reactor, could deliver during the same period. So, once you increase radiation from about 1,000 W/m2 (impossible to achieve on earth due to weather and other realities) to 1,400 W/m2, 24/7, no weather, etc. The scenario quickly becomes vastly different.

Geothermal is very interesting, thanks for the ticker, I'll look into it.

All that said, I think the US (and Europe, if they care to survive) needs to have, as a top national priority, a massive program for the construction of nuclear power plants.

A few years ago, I ran through an analysis of power generation needs to convert our entire fleet of vehicles to electric power. That required at least doubling our power generation capacity. This was the equivalent of having to build 1,200 nuclear power plants, each producing 1 GW 24/7.

This is impossible to achieve with solar, or wind, or the combination.

So, even if we decide not to care one bit about AI data centers, we need to double our power generation capacity (and infrastructure to deliver it).

Add AI data centers to that equation and the number could easily climb another 600 nuclear power plants.

Here's the problem that a naïve conversation with an LLM will not uncover: Humanity and Politics.

Could we embark on such a massive energy-generation project? Yes, absolutely.

Is it realistic? Nope. Not even close.

In other words, we can do it but it is impossible...which sounds like an oxymoron until you consider that we have lots of examples of projects that are absolutely plausible that, once they contact political and government reality, quickly become impossible. The best-worst example I always grab is the California high speed train disaster.

LLMs are a national security issue? Why is China releasing their models then?

The question is not whether you should have datacenters, or the most advanced chips, or the ability to build the most capacity. But do we need all this now? Will there be enough demand? People want to make profits form this thing, and what's being pointed out is that maybe there won't be enough demand to generate profits for all this investment.