in part because I don’t have much to say on this issue that Gary Marcus hasn’t already said.
It would be interesting to know which particular arguments made by Gary Marcus you agree with, and how you think they relate to arguments about timelines.
In this preliminary doc, it seems like most the disagreement is driven by saying there is a 99% probability that training a human-level AI would take more than 10,000x more lifetimes than AlphaZero took games of go (while I’d be at more like 50%, and have maybe 5-10% chance that it will take many fewer lifetimes). Section 2.0.2 admits this is mostly guesswork, but ends up very confident the number isn’t small. It’s not clear where that particular number comes from, the only evidence gestured at is “the input is a lot bigger, so it will take a lot more lifetimes” which doesn’t seem to agree with our experience so far or have much conceptual justification. (I guess the point is that the space of functions is much bigger? but if comparing the size of the space of functions, why not directly count parameters?) And why is this a lower bound?
Overall this seems like a place you disagree confidently with many people who entertain shorter timelines, and it seems unrelated to anything Gary Marcus says.
I agree with essentially all of the criticisms of deep learning in this paper, and I think these are most relevant for AGI:
Deep learning is data hungry, reflecting poor abstraction learning
Deep learning doesn’t transfer well across domains
Deep learning doesn’t handle hierarchical structure
Deep learning has trouble with logical inference
Deep learning doesn’t learn causal structure
Deep learning presumes a stable world
Deep learning requires problem-specific engineering
Together (and individually), these are good reasons to expect “general strategic action in the world on a ~1-month timescale” to be a much harder domain for deep learning to learn how to act in than “play a single game of Go”, hence the problem difficulty factor.
It would be interesting to know which particular arguments made by Gary Marcus you agree with, and how you think they relate to arguments about timelines.
In this preliminary doc, it seems like most the disagreement is driven by saying there is a 99% probability that training a human-level AI would take more than 10,000x more lifetimes than AlphaZero took games of go (while I’d be at more like 50%, and have maybe 5-10% chance that it will take many fewer lifetimes). Section 2.0.2 admits this is mostly guesswork, but ends up very confident the number isn’t small. It’s not clear where that particular number comes from, the only evidence gestured at is “the input is a lot bigger, so it will take a lot more lifetimes” which doesn’t seem to agree with our experience so far or have much conceptual justification. (I guess the point is that the space of functions is much bigger? but if comparing the size of the space of functions, why not directly count parameters?) And why is this a lower bound?
Overall this seems like a place you disagree confidently with many people who entertain shorter timelines, and it seems unrelated to anything Gary Marcus says.
I agree with essentially all of the criticisms of deep learning in this paper, and I think these are most relevant for AGI:
Deep learning is data hungry, reflecting poor abstraction learning
Deep learning doesn’t transfer well across domains
Deep learning doesn’t handle hierarchical structure
Deep learning has trouble with logical inference
Deep learning doesn’t learn causal structure
Deep learning presumes a stable world
Deep learning requires problem-specific engineering
Together (and individually), these are good reasons to expect “general strategic action in the world on a ~1-month timescale” to be a much harder domain for deep learning to learn how to act in than “play a single game of Go”, hence the problem difficulty factor.