One fundamental challenge to me is that if each training run because more and more expensive, the time it takes it to learn what works/doesn't work widens. Half a billion dollars for training a model is already nuts, but if it takes 100 iterations to perfect it, you've cumulatively spent 50 billion dollars... Smaller models may actually be where rapid innovation continues simply because of tighter feedback loops. O3…
AGI is the Sisyphean task of our age. We’ll push this boulder up the mountain because we have to, even if it kills us.
I think you're both right and wrong. You're right that capitalism has become a paperclip machine, but capitalism also wants AI so it can cheaply and at scale replace the human components of the machine with something that has more work capacity for fewer demands.
The problem is that the people in power will want to maintain the status quo. So the end of human labor won't naturally result in UBI – or any kind of welfare – to compensate for the loss of income, let alone afford any social mobility. But wealthy people will be able to leverage AGI to defend themselves from any uprising by the plebs. We're too busy trying to make humans irrelevant, but not asking what exactly we do…
The problem is that the "we" who are busy trying to make humans irrelevant seem to be completely unconcerned with the effects on the "we" who will be superfluous afterwards.
O3 has demonstrated that OpenAI needs 1,000,000% more inference time compute to score 50% higher on benchmarks. If O3-High costs about $350k an hour to operate, that would mean making O4 score 50% higher would cost $3.5B (!!!) an hour. That scaling wall.
I used to run a lot of monte carlo simulations where the error is proportional to the inverse square root. There was a huge advantage of running for an hour vs a few minutes, but you hit the diminishing returns depressingly quickly. It would not surprise me at all if llms end up having similar scaling properties.
Yeah, any situation you need O(n^2) runtime to obtain n bits of output (or bits of accuracy, in the Monre Carlo case) is pure pain. At every point, it's still within your means to double the amount of output (by running it 3x longer than you have so far), but it gradually becomes more and more painful, instead of there being a single point where you can call it off.
One fundamental challenge to me is that if each training run because more and more expensive, the time it takes it to learn what works/doesn't work widens. Half a billion dollars for training a model is already nuts, but if it takes 100 iterations to perfect it, you've cumulatively spent 50 billion dollars... Smaller models may actually be where rapid innovation continues simply because of tighter feedback loops. O3…
AGI is the Sisyphean task of our age. We’ll push this boulder up the mountain because we have to, even if it kills us.
I think you're both right and wrong. You're right that capitalism has become a paperclip machine, but capitalism also wants AI so it can cheaply and at scale replace the human components of the machine with something that has more work capacity for fewer demands.
The problem is that the people in power will want to maintain the status quo. So the end of human labor won't naturally result in UBI – or any kind of welfare – to compensate for the loss of income, let alone afford any social mobility. But wealthy people will be able to leverage AGI to defend themselves from any uprising by the plebs. We're too busy trying to make humans irrelevant, but not asking what exactly we do…
Machines can’t have fun for us. They can’t dance to a beat, they can’t experience altered states of mind. They can’t create a sense of belonging through culture and ritual. Yes we have lost a lot in the last 100 years but there are still pockets of resistance that carry old knowledge that “we the people” will be glad of in the coming century.
> And the results of the project, dubbed Arrakis, indicated that creating GPT-5 wouldn’t go as smoothly as hoped. Quite the hubris to name the project after the desert planet of Dune, where multiple royal houses met their ruin.
And the other theme in Dune is how artificial intelligence essentially fubar’d civilization. (The Butlerian Jihad)
"There is no evidence that LLMs are the roadmap to AGI." - There's plenty of evidence. What do you think the last few years have been all about? Hell, GPT-4 would already have qualified as AGI about a decade ago.
> What do you think the last few years have been all about? Next token language-based predictors with no more intelligence than brute force GIGO which parrot existing human intelligence captured as text/audio and fed in the form of input data. 4o agrees: "What you are describing is a language model or next-token predictor that operates solely as a computational system without inherent intelligence or understanding. T…
o1 points out this is mostly about “if submarines swim”.