I'm reminded of what the Benetton F1 team did when rules specifically prevented teams from using wheel speed and gear information to run traction control systems. Instead they used a combination of incoming air stream pressure (similar to how an aircraft's pitot-static system works) in combination with some preloaded per-event data and track position information to make another traction control system that did the same job.

Information about past leases and public data from competing landlords might still be enough to do the job.

An LLM may be an ingredient, but the current commercial approach of trying to just build an LLM so big that it magically becomes an AGI somehow (or I think the idea may be to make a stochastic parrot good enough to be hard to distinguish from an AGI) sure doesn't look anything like a path, especially when the returns are diminishing so hard.

Totally agree. Today's "AI" companies would like their investors to believe there's a path though.

Clearly we're misunderstanding each other. I was saying that investing in the bubbled asset was folly. I think the AI industry will continue to exist after the bubble pops but at a size no larger than the database industry today. So not zero value, but a small fraction of what it is currently.

I do think that the amount of money being invested in AI training for the improvements being produced is an absurd waste. They're spending larger and larger sums of money to produce rapidly vanishing improvements that customers have so far never shown an interest in paying enough to turn a profit with.

So in other words, because chewy.com exists today it would've been smart to invest into pets.com at the height of its value?

This is going to be such a disastrous investment it's going to make Solyndra look like an insignificant whoopsie in comparison. While Chinese product dumping efforts can be hard to foresee, the obviousness and severity of the AI bubble has been on public display for anyone who cares to look for months now. And there's the potential to sink far more money into it. The winner of the AI race is going to be whoever wastes the least money on this folly, and the US looks set for a massive and easily avoidable loss now.

And let's not forget the end goal of this. If someone were to win this race in the fictional imagined scenario where AI didn't hit the core of Diminishing Returns Planet around ChatGPT 4 and there was some kind of path from LLM tech to AGI, the end result would be a technology that augments/replaces labor (same thing, don't be fooled by your boss) in a world dominated by an economic system where most people are workers who need to be able to find buyers for their labor. What could possibly go wrong with that?

Wouldn't be the first time I needed to change a default config to work around NSA / Five Eyes fuckery, won't be the last:

https://blog.stribik.technolog...

Even with antifreeze, attempts to thaw bodies prepared that way have resulted in massive widespread organ cracking.

Could also run some World Community Grid tasks on BOINC, that's what I do.

But if you just want to heat your house nothing beats a heat pump for efficiency.

I don't think we should expect to see steady progress toward a climate solution, I think it's going to happen quite suddenly after some combination of the technology to do it getting cheap enough and the climate producing enough "shit's getting real" moments for a large fraction of the first-world population. At that point either we'll get our asses in gear and set up oceanic and atmospheric carbon sequestration megastructures all over the planet with maybe a little SRM sprinkled on top, or we'll be too distracted or impoverished to do anything about it for some silly reason or another and all the disastrous predictions will come true because this is a problem we're collectively too stupid as a species to solve. That's a real possibility.

They could've instead named this thing "Wut is sequential access and why is it bad!?!?"

It's also worth noting that NAFTA obliterated Mexico's smaller farmers which massively ramped up illegal migration from Mexico to the US.

I think the actual winners of this AI bubble will be whoever wastes the least money on this folly.

This.

I'm also surprised he didn't mention the housing bubble. Just because it's been running for decades and has been intentionally stabilized in an inflated state doesn't make it less of a bubble.

This wouldn't be the first project Microsoft killed with their own greed and it won't be the last.