It has the lifetime anchor on one end, and the evolution anchor on the other end, and almost all probability mass is in between, and then it has a parameter for how that mass shifts leftwards over time as new ideas come along.
This reminds me—so what I would really like to see is a distillation of Ajeya’s model into a post (with some pretty pictures) that more deeply explores this ‘space in between’, deeply grounded in deep knowledge of both DL and neuroscience, that tracks the evolutionary history of both, pairing and comparing milestones—starting with say perceptrons and jellyfish (or insert early equivalent) moving all the way up to current SOTA agents vs (insert animals here).
That would allow us to establish a rough technological evolution / biological evolution speedup curve and project forward—and most likely substantiate the case for short timelines (I predict).
This reminds me—so what I would really like to see is a distillation of Ajeya’s model into a post (with some pretty pictures) that more deeply explores this ‘space in between’, deeply grounded in deep knowledge of both DL and neuroscience, that tracks the evolutionary history of both, pairing and comparing milestones—starting with say perceptrons and jellyfish (or insert early equivalent) moving all the way up to current SOTA agents vs (insert animals here).
That would allow us to establish a rough technological evolution / biological evolution speedup curve and project forward—and most likely substantiate the case for short timelines (I predict).