When will there be AI superintelligence?
-
Thanks everyone! I have to admit that I'm not sure. Here's my best guess.
One way to guess at whether a system can handle intelligence is to estimate the complexity of that system. The human brain has 86 billion neurons and about 100 trillion synapses, the connections between the neurons. So, we could make a claim that systems won't become intelligent until they're at least that complex. This is Ray Kurzweil's estimation method.
Currently the biggest LLMs have about 0.5 trillion parameters -- the weights between nodes in the neural network. They're simpler than synapses, but they serve about the same function.
-
Currently the biggest LLMs have about 0.5 trillion parameters -- the weights between nodes in the neural network. They're simpler than synapses, but they serve about the same function.
Models are growing at about 2-3x per year. If it's possible for them to keep growing at that rate, it will be about 8-9 years before they get to 100 trillion.
-
Currently the biggest LLMs have about 0.5 trillion parameters -- the weights between nodes in the neural network. They're simpler than synapses, but they serve about the same function.
@evan "they serve about the same function"
Citation? Naming them neural nodes doesn't count
-
Thanks everyone! I have to admit that I'm not sure. Here's my best guess.
One way to guess at whether a system can handle intelligence is to estimate the complexity of that system. The human brain has 86 billion neurons and about 100 trillion synapses, the connections between the neurons. So, we could make a claim that systems won't become intelligent until they're at least that complex. This is Ray Kurzweil's estimation method.
@evan 27b models are now as good as huge frontier models of a year ago. I don't think the idea of just scaling will alter the fundamental issues with LLM ML.
-
Models are growing at about 2-3x per year. If it's possible for them to keep growing at that rate, it will be about 8-9 years before they get to 100 trillion.
One way to define "superintelligence" is about 100x greater than human intelligence. That would be another 7-8 years.
-
@evan this is not just a casual interest for you. You crafted the questionnaire and your responses to it. Would I be correct in saying this?
-
One way to define "superintelligence" is about 100x greater than human intelligence. That would be another 7-8 years.
So, that's about 15 years. Let's say give or take 5 years. And assuming that it's even possible with LLM technology. But, if I were going to put money on it, I'd say around 2040.
-
So, that's about 15 years. Let's say give or take 5 years. And assuming that it's even possible with LLM technology. But, if I were going to put money on it, I'd say around 2040.
A lot of the replies say that it will never happen. My guess is that some significant portion of the majority of people who answered 2060 or later mean "never" or "in the far future".
-
@evan you must be compiling these things for a reason?
-
@evan do you have any findings you'd like to share?
-
A lot of the replies say that it will never happen. My guess is that some significant portion of the majority of people who answered 2060 or later mean "never" or "in the far future".
I think it's possible that it's never. We only have one example of known intelligence, and a lot of interesting close calls, including not only some pretty smart mammals, but also eusocial insects as collective intelligences. We just don't know enough about intelligence to say what kinds of systems it can arise in, and what kind of systems it can't.
-
@evan or are your smuggled premises the only point?
-
I think it's possible that it's never. We only have one example of known intelligence, and a lot of interesting close calls, including not only some pretty smart mammals, but also eusocial insects as collective intelligences. We just don't know enough about intelligence to say what kinds of systems it can arise in, and what kind of systems it can't.
@evan I think our definitions of "sentience" and "intelligence" are going to radically change radically soon..
-
I think it's possible that it's never. We only have one example of known intelligence, and a lot of interesting close calls, including not only some pretty smart mammals, but also eusocial insects as collective intelligences. We just don't know enough about intelligence to say what kinds of systems it can arise in, and what kind of systems it can't.
I don't think we know for sure a lot about intelligence. We don't know that it can be implemented on silicon chips, but we also don't know if it can't. We don't know if it has to work like brains do, or if it could work in entirely different ways. We don't even know exactly how our own minds arise from our brains. We don't know if we can replicate intelligence without understanding our own first.
-
I don't think we know for sure a lot about intelligence. We don't know that it can be implemented on silicon chips, but we also don't know if it can't. We don't know if it has to work like brains do, or if it could work in entirely different ways. We don't even know exactly how our own minds arise from our brains. We don't know if we can replicate intelligence without understanding our own first.
I'm pretty dubious about claims that superintelligence is so immanent, and so dangerous, that LLM technology should be restricted to only a few trusted companies and the US government. That sounds like some self-serving bullshit by companies that want to consolidate economic power by partnering with a protectionist government.
-
So, that's about 15 years. Let's say give or take 5 years. And assuming that it's even possible with LLM technology. But, if I were going to put money on it, I'd say around 2040.
@evan I guarantee you a text-predictor cannot become intelligent.
-
I'm pretty dubious about claims that superintelligence is so immanent, and so dangerous, that LLM technology should be restricted to only a few trusted companies and the US government. That sounds like some self-serving bullshit by companies that want to consolidate economic power by partnering with a protectionist government.
All that said, AI is in a very different place than it was at the beginning of my working career, during the "AI Winter" of the early 90s. Deep learning really changed a lot. We use AI techniques now a lot more, for a lot more areas of applicability. And LLMs are really impressive.
-
All that said, AI is in a very different place than it was at the beginning of my working career, during the "AI Winter" of the early 90s. Deep learning really changed a lot. We use AI techniques now a lot more, for a lot more areas of applicability. And LLMs are really impressive.
There have been previous generations of AI technology that hit walls -- the planning systems of the 1950s, the expert systems of the 1980s. It's possible that LLMs will hit some kind of wall -- that it won't be possible to get to superintelligence with that method.
-
There have been previous generations of AI technology that hit walls -- the planning systems of the 1950s, the expert systems of the 1980s. It's possible that LLMs will hit some kind of wall -- that it won't be possible to get to superintelligence with that method.
Oh, and we're going into a dangerous century. We might not have the social cohesion, much less the resources, to maintain huge AI labs through another 15 levels of doubling complexity.
-
Oh, and we're going into a dangerous century. We might not have the social cohesion, much less the resources, to maintain huge AI labs through another 15 levels of doubling complexity.
All that said, my gut is still pretty good with that 2040 date.
