He's well compensated because he owns a lot of the company, much at much cheaper valuations, right?
For all the complaints about him you can't argue he hasn't delivered a lot of value. Lots of corp CEOs make more with falling company performance and valuation, because they aren't really paid for performance.
The conclusion I've come to is that AI is still revolutionary, and the buildout is probably necessary to some degree, but it is way overestimating the demand and pricing power providers will have.
SW can use a ton of AI. Lots of other applications can't, and they don't need the super smart frontier versions either. We just don't need all that much compute to serve AI for these applications, and people won't need to pay the margins necessary to get access to that AI. SW alone is not going to be enough to pay for all the investment.
I don't buy brand loyalty either. If I can pay 1$/m for the same thing from someone as I pay OpenAI or Anthropic 10$/m for, I'll do that, which almost zero barrier to entry.
The buildout is paying stratospheric prices on the assumption they can charge that much on the other end, but if demand tapers off (or model efficiency doubles again), prices will start falling. Everyone talks about Jevon's, but I don't really see why I'd spend more on AI than I need to do the things I want. I already run out of stuff to ask it to do on a basic subscription, I hardly need more.
The crux is this: I can spend x for y results. If I spend more, I probably don't get 10x for 10y results - I probably get, say, 2y results. So on that equilibrium I just don't spend all that much more because the reward for all that extra spend is not worth it. And if models get better this tradeoff starts to look worse, since I get more of the super-effective benefits for cheaper. Now I'm spending 0.1x for y results, and spending even x for 2y might not make sense, let alone 10x for 3y.
The problem is, that its not just AI for a niche thing, its AI together with compute.
People are already using AI to build them some things with CAD. Electronics too. But now what does AI need to be very good in it without an expert? Simulation and Optimizers. This again needs more compute.
And the agentic layer will be added to all big products out there. Everything you will do will consume AI tokens
To be fair, Jevons suggests that as prices fall, novel use-cases are unlocked. The things you or others currently do with AI might not need frontier models or loads of tokens, but if prices fall or capabilities increase, new applications might arise to exploit that.
Yes, this reasoning has been used to justify the absolutely gigantic amount of spending going into AI right now.
I'm not convinced it's true. It just seems to be taken for granted that demand for intelligence is unlimited and that its cost limited, which just doesn't seem true to me.
I don't understand how anyone can call these AGI. I've always had the view that AGI == Independent human-level intelligence.
At some point someone moved the goalposts for this to include LLMs that emphatically are not that. I think it was Altman when they came out with Astra?
Please do a youtube series where you ask random people on the street some questions and have them go against Opus 5.5. The model is definitely smarter than me in many ways- it would obliterate me in trivia, could write a poem in iambic pentameter before I could remember or look up what exactly the rules for it are, would probably dominate me at chess or football strategy, would do a better job at taxes, and for sure writes better code outside my small little slice of specialty. It definitely parses and diagnoses logs better than me unless I wrote all the loggging statements myself.
I have to ask, what humans do you know that have more general intelligence? This isn't a dream team where you can say "Magnus Carlson can beat it at chess, and Linus can write better C code", this is a comparison across all domains and levels of intelligence. Even its IQ is estimated at 141, which is seriously higher than the average human. The average human's reading level is atrocious. The average human cannot write a standard 5 paragraph essay about a random topic.
If I choose an arbitrary video game, board game or whatever can Opus play it?
My understanding is that models will often still forget games rules etc unless you design a harness specifically to assist them and prevent them going off the rails. I was going to use chess as an example but it sounds like they are good enough to do that now, albeit I'm not sure if they have a special harness for it.
It's not about how "smart" they are, or how good they are at recall. It's about whether I can teach a model to understand and execute an arbitrary, general task regardless of whether they have been trained on that task before.
Disregarding the above, the fact models still require a human prompter/decision-maker is defacto proof they aren't AGI in my view. If they were humans would be obsolete in decision-making.
I guess im doing a terrible job of explaining my view.
If these were AGI we wouldn't be talking about them helping engineers, engineer jobs would just not exist. The AI would be superior in every way that mattered. That's not the case yet, so IMO we don't have AGI.
they aren't as intelligent as a human, but they can throw shit at the wall faster, and get something that sticks faster. they are superhuman in areas that require trial and error, like hacking.
everyone stopped talking about the Turing test once ChatGPT came out and broke right through it while still being very very dumb. AGI is just too blurry to define.
I hate these glib comments that don't remotely reflect reality.
Go look at the goods people got in countries that didn't practice capitalism. They were worse in almost every way, more expensive if you could get them and often the things people wanted were completely unavailable because there was no market incentive informing production.
In my experience, this sort of comment is usually followed up by some personal preference of the commenter. Great. I don't want the same thing, so I guess something has to mediate the resource allocation and effort of society towards providing that want/need at a certain quality/price.. and guess how we do that?
Building all of the systems to train and serve our models, internal tools and infrastructure, (lots and lots of) research, our product and API surfaces, and more. We have a lot of problems that we can solve with code and there are all sorts of new projects spinning up every week.
Well, you can extract all the value created using your tools and inventions, because then no one would have any incentive to use them to create value.
They try to get around this by charging you more after you've invested in your tools, but even then you can only go so far (though it's astonishing how far the big guys have managed to get regardless)
If anything, the core job of most software is to reliably and repeatedly do the thing a user wants. Bugs outside eof that are OK, but a software that doesn't Do The Thing is useless.
LLMs increasing bug density in exchange for more features is likely a terrible tradeoff for most software people use IMO.
For all the complaints about him you can't argue he hasn't delivered a lot of value. Lots of corp CEOs make more with falling company performance and valuation, because they aren't really paid for performance.
reply