181–190 of 208 posts
Yann is a nobody at Meta (lol at calling FAIR a top lab at this point, it’s not even the top org in Meta for AI) and keeps trying to claim credit by association for Meta LLM success.
Earlier quoted context omitted.
Before the era of quantum physics, the idea of a weapon like fission or fusion weapons would've been inconceivable without resorting to magical thinking. Your argument, and Yann's, is that AGI, or what you call AGI, is a kind of quasi-intelligent golem that, despite being generally intelligent, doesn't have human-level intelligence. Your claim that it will never be human-equivalent, much less trans-human/ASI, is buil…
What is human-level intelligence is a normative and philosophical question, which means it will not be resolved in some objective manner. There's no requirements for us to know, there's only requirements for us to propose in the form of norms. The only approach we will be able to claim objectively can produce systems with human-level intelligence is procreation.
Of course over 100 years later we know it's not only possible but manned flight is far more capable than natural flight in almost every metric.
Earlier quoted context omitted.
I've read a fair bit of LeCunn and frankly have been very unimpressed. He is, as far as I can tell, the only big AI name who doesn't think that if we somehow managed to make an actual superintelligent AI it wouldn't be dangerous.
His reasoning is that good guy’s AI will counteract bad guy’s ones or a an AI that has developed his own objective against humanity if that happens, which is I think a reasonable argument.
Earlier quoted context omitted.
> Anything closed source and having regulatory oversight is useless Source isn't even the problem, unless you're a billionaire, you can't afford to train the model. This isn't the wheel, the printing press, the PC, or the public internet, which created opportunities for everyone . This is pay-to-play that is only affordable to the nation-state the mega-corporation, and the latter might let you play around as a digita…
The model is the instance because the instance is what is being used at inference time. The training simply produces some set of weights. Open source would mean someone could see and run the model code locally/independently to create an instance. For an LLM, this is much less insurmountable of an issue in terms of resources needed.
We are still answering questions on LLM/AI scaling. Ya, you might have your cottage industry AI giving out answers on the trickle of information you feed it, but will that even be comparable to one being fed megawatts of power with terabytes of data per second flowing into its databanks?
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-...
"Elon Musk doesn't know how to code" is my fav. It's as if coding is some sort of mythical and rare skill. A new grad who just learn to code a year ago is able to intern at FAANG, but Elon Musk cannot possibly learn how to code for some reason.
Earlier quoted context omitted.
Show me LLM that has reached a level of automony and self sufficiency of a cat (Yann's example) or even of a simple beattle. LLMs are tools in amplifying individual human intelligence, 100% of automony and will come from their user (human). Also, even as tools they have fundamental limitations which stem from their autoregressive nature
Before the era of quantum physics, the idea of a weapon like fission or fusion weapons would've been inconceivable without resorting to magical thinking. Your argument, and Yann's, is that AGI, or what you call AGI, is a kind of quasi-intelligent golem that, despite being generally intelligent, doesn't have human-level intelligence. Your claim that it will never be human-equivalent, much less trans-human/ASI, is buil…
Earlier quoted context omitted.
Before the era of quantum physics, the idea of a weapon like fission or fusion weapons would've been inconceivable without resorting to magical thinking. Your argument, and Yann's, is that AGI, or what you call AGI, is a kind of quasi-intelligent golem that, despite being generally intelligent, doesn't have human-level intelligence. Your claim that it will never be human-equivalent, much less trans-human/ASI, is buil…
This is not about intelligence, it's about agency. Text completion generators don't have agency no matter how big they get.
Earlier quoted context omitted.
Before the era of quantum physics, the idea of a weapon like fission or fusion weapons would've been inconceivable without resorting to magical thinking. Your argument, and Yann's, is that AGI, or what you call AGI, is a kind of quasi-intelligent golem that, despite being generally intelligent, doesn't have human-level intelligence. Your claim that it will never be human-equivalent, much less trans-human/ASI, is buil…
This is not about intelligence, it's about agency. Text completion generators don't have agency no matter how big they get.
https://promptengineering.org/what-are-large-language-model-...
Earlier quoted context omitted.
No one should respect that blowhard. Please point me to the specifically accurate predictions he has made so far.
Eliezer made early predictions that we would solve the protein folding problem when effectively everyone else said we either wouldn't solve it or it was way off. If I remember right he was making those claims around 2005 or so and predicting wed have it solved in something like 30 years. We solved it in 2017.