Why did General Nakasone join OpenAI? Because THIS. There is a TON of work to do!
(twitter.com)
🐸 WAG THE FROG 🐸
You're viewing a single comment thread. View all comments, or full comment thread.
Comments (74)
sorted by:
Well, I think you're oversimplifying things here a little bit. Although he may have a conflict of interest, it's clear from his work that he is in close collaboration and very closely exposed to the behind the scenes realities of these projects. He's not merely extrapolating the progress of AI in a linear way, in fact, the progression from gpt2 to gpt4 was itself logarithmic (as is Moore's Law, BTW). Both types of extrapolations can track real world trends given sufficiently robust real world tracking data.
To clarify, his projections of growth rates into the future are linear (not the underlying growth rates themselves). In other words, he projects that the rate of growth seen in the past will be the same rate of growth going forward in the future (that is the linear part). This rarely if ever happens in reality.
The other big problem he has is that he talks as if the current machine learning models already have the equivalent of human intelligence of varying levels, when in fact they have literally zero intelligence. This is probably why he is so far off regarding his timeline to achieve AGI. ChatGPT4 is nowhere near true AGI. Not even close.
Now factor in that he just started a firm to invest in AGI. If that is 30 to 50 years in the future, it would be difficult to encourage people to invest in his new firm... but if AGI is magically "only a few short years away"... I think you can figure out his motives for yourself.
Back in the DotCom bubble, people fervently believed that all kinds of crazy shit was true... until the whole thing crashed and burned, and reality bitch-slapped everyone into the next decade. Good times...