To the extent Altman pretends it is a physics problem by gesturing at nuclear fusion, he’s full of shit. The problem of energy is as solved as it will likely be for many decades. We know how to power these data centers and have a flexible menu of options. The problem is building it out.

No, I am annoyed by people (including Altman and other grifters!) acting like physics is relevant. Nothing about this is a physics problem. It’s an infrastructure development problem.

I feel like I'm losing my mind. Altman is saying over and over "we can't generate the obscene amounts of power required by generative AI" and everyone's pointing and laughing and saying "he thinks Dyson Spheres are real! he thinks you can make electrons!"

That is almost surely your problem lol. I don’t think about physics at all unless its laws intrude upon my life. And utility companies (and now apparently AI companies) know many investors are the same way and tailor their message to the lowest common denominator (me).

I agree that’s bad. But I don’t think you can intelligently critique AI without confronting their economic vision and breaking it down. “Har har he doesn’t know physics” and “lol jargon bad,” despite ample contextual information, are not that.

I have repeatedly said he is an asshole. He’s just not as dumb as people are insinuating, and such insinuations don’t do any credit to their own intelligence.

As they take on more debt, that could change, but the current risk is still tolerable. People just need to expect much lower yield going forward and actually diversify to counter that.

As for everyday investors getting scammed, I don’t think we’re there yet. The companies to which investors are actually exposed are nowhere near as risky as the model companies and other emerging players.

I don’t think Altman saying “electrons” instead of “power” is going to change anything. The real issue is the federal government and whether they can extract some bullshit support from this corrupt administration.

Idk man. People On Here have gotten the message loud and clear, and it’s a political issue that is increasingly moving the needle irl. Lots of state utility commissions are studying the issue now (though these body move like molasses as a rule).

I’m sure you understand the technology better than Altman. What is missing is his fanciful economic model based on scarcity and a winner-take-all vision for AI development. If you accept that (which to be clear I don’t), then you’ll bear any infrastructure development costs to win the race.

When people have made those points, I have agreed with them. But many of the comments have remained focused on the technical language and not the substance.

Because that isn’t what they were saying.

Altman likes to compare compute to oil, i.e., he sees it as a strategic resource that should be hoarded to the greatest extent possible. Seems intuitive enough, but there are other ways to describe this concept.

I am annoyed that the critique is “lol dumb science words” instead of actually addressing the economic point he is making. It is a fact that their ability to scale is constrained by energy availability, and they’re obviously looking for ways to socialize the costs of solving that problem.

One could even argue that by spreading the idea that he was talking about literal electrons, which he clearly wasn't, YOU are spreading disinformation.

But you are going to subtweet me with something childish for saying that, aren't you?

He pretty much said just that, which is why it’s odd to me that people were so focused on his scientific “illiteracy” instead of the substantive economic issue of the energy math not mathing.

futurism.com/openai-altma...

Again, I agree with you that premise is flawed, but that’s their whole pitch. Anyone who understands what is happening in this industry knows that and knows exactly what the quoted exchange means. Pretending that it is mysterious when it is not is pointless and detracts from the valid critique.

The same reason companies say “optimize headcount” instead of “reduce headcount to the greatest extent possible.” The C-suite hates long emails, overly detailed investor materials, etc.