121–130 of 208 posts
I dunno all this talk about OpenAI lately really feels more and more like hype to me. What they've accomplished so far is really impressive. But until I see AGI for real I won't hold my breath. Just feels to me like some expert/wizard level PR is going on right now.
One thing that makes no sense about Sam Altman hiding the Q* progress from the board is that Ilya Sutskever was part of the board too. As he is the technical lead and Altman isn't technical at all, how would it be a secret from Sutskever? NB: I know nothing about these people and don't track it, but my consistency check failed on this claim based on the flood of news over the past week.
Where did the myth of Sam "not being technical at all" come from? Patrick Collison seemed satisfied with Sam's tech chops [1]. If that isn't enough of a vouch, god help the rest of us. [1] https://www.quora.com/Is-Sam-Altman-highly-technical-Has-he-...
Also, I wouldn't call it a myth. I would call it 'questions around his technical knowledge'.
The best way to answer those questions would be for direct evidence of his technical ability to be shown, not references from mates (just as any interview for a tech position requires direct evidence).
> One of the main challenges to improve LLM reliability is to replace Auto-Regressive token prediction with planning. Anyone have a good source for learning more about this topic?
https://arxiv.org/pdf/2305.20050.pdf
It’s different from the link the other user posted.
This one is specifically about the Q* process
I dunno all this talk about OpenAI lately really feels more and more like hype to me. What they've accomplished so far is really impressive. But until I see AGI for real I won't hold my breath. Just feels to me like some expert/wizard level PR is going on right now.
He’s been consistently voicing realistic statements with regards to AGI (he doesn’t believe it’s a thing), AI safety and doomerism, limitations of LLMs, etc.
I dunno all this talk about OpenAI lately really feels more and more like hype to me. What they've accomplished so far is really impressive. But until I see AGI for real I won't hold my breath. Just feels to me like some expert/wizard level PR is going on right now.
If agi is possible, whomever creates it is going to do their damnest to make sure no one else finds out for as long as possible.
When we have AGI, we will find it out pretty quickly since whoever gets there first wants the most money and control they can.
Earlier quoted context omitted.
LeCun at least knows a lot about AI. Most of the Q* stuff is coming from people who know almost nothing.
Seems to be a common theme with stuff starting with Q...
I could not imagine the same amount of searches and guesswork had they called it "P*" instead.
It seems like Q inherently has some mystery to it. Maybe in that is the least commonly used letter and it rarely is used separately from "Qu"
So Q* is just A* for neural networks?
So Q* is just A* for neural networks?
I haven't heard any real details of what Q* is supposed to be; I assumed it's related to reinforcement learning. I guess it's possible the heuristic in A* was replaced with a neural network value function. I'm not sure how that could lead to a breakthrough in language models though.
Q star would mean optimal Q. If that is even what the star is referring to. There is still no way to know if the Q is the quality factor, query value, question, something else, or just a placeholder name that doesn't reference its structure at all
Earlier quoted context omitted.
The point is moot anyway. If someone can build it someone will. Laws and impending Doom or not. It's probably going to be better if it's us than Russia or China.
That's a reductive argument, and won't work with me. What we have is tantamount to an arms race, and trying to suppress another country's development of tech that could be used against them is a thing. We have already restricted access to our microchips to China specifically and Russia through broad sanctions. https://www.nytimes.com/2023/10/17/business/economy/ai-chips... Russia isn't poised for the current crop of…
What they can do is ensure that the sanctioned entities are always at least a generation behind. That's not especially helpful in the context of "anti-doomers will be wrong exactly once" unfortunately.