Happy Sunday and welcome to InsideAI! I'm Rob May, CTO at Dianthus. We buy ecommerce companies and use our proprietary AI/ML software to grow them faster and improve the bottom line. If you know an ecommerce brand that is $2M - $20M in revenue and interested in selling, send them my way. I'm also a very active AI angel investor, and always on the lookout for new AI startups.
Today I want to talk about predictions of when we reach a more general version of artificial intelligence, similar to a human brain, and ask what we've learned. There have been a few approaches to this over the years. One that I was a big fan of was this 2015 WaitButWhy piece on the AI revolution. The argument in this piece is that AI progress is doubling and we are expecting a linear trend, but that doubling will explode the AI capabilities of machines sooner than we thought. I admit that I was a big fan of this argument, but it increasingly looks incorrect. While it is possible that this is true, and that we are still just in the early stages of the trend, it increasingly looks like the marginal gains from existing approaches to AI are decling and won't get us to general AI.
The other big prediction about when we get there is Ray Kurzweil's extrapolation of computing power, noting that next year, in 2023, the amount of compute you can buy for $1000 will surpass the compute available in the human brain, bringing us close go general AI. Of course, that only works if the key to AI is raw compute power. It increasingly looks like that may be wrong.
I was a fan of Kurzweil's approach earlier in my career but moved away from it in favor of other options. I've sensed moved on from those options as well as most things seem to be wrong.
I recently read the book Spike, about a tour of what happens in the brain in a short 2 second window when it is making an evaluation. It appears we are learning that the computational practices of the brain are quite different than we thought, and quite different from digital computers, and thus we aren't as close to replicating them as we thought.
Now there are a few theories about what can happen next. One that some AI experts believe is that robots will be the key to general AI, because intelligence can't be just floating around on a chip, it needs to be grounded in the real world with sensors, and thus robotics can take us there. Another theory is that prediction, and the constant comparions of prediction to reality are what drive intelligence. The most popular version of this theory is probably Jeff Hawkins' On Intelligence.
But I favor the approach that my friend Roman Yampolskiy, has been harping on for the past 7 or 8 years. He says "we don't know." His point is that we don't yet know enough to know if it's a compute problem, a data problem, an algorithm problem, or something else. It's very possible that we finally get enough data, or enough compute, and we crack general artificial intelligence. It's also possible it's a discontinuity - a jump that doesn't tie to anything we've been doing so far, which would say that whatever patterns we are seeing aren't indiciative of the path to general AI. It's possible that someday a lab, or a company, or a college kid sitting in a dorm room, comes up with a new learning algorithm that is THE key breakthrough we need to push AI forward. We just don't know.
For now, based on the trends I'm seeing, it doesn't look like continuing any of them at higher levels will get us there, so I like the discontinuous jump theory.
Thanks for reading.
@robmay
Comments
Post a Comment