Turns out that majoring in gradient descent studies is a poor ersatz for a working knowledge of the history of AI progress, neuropsychology, developmental psychology, and mathematics
Show this thread
One of the saddest things about the field of AI is how ahistorical and siloed it is. Some of the key debates of our day (e.g. innate priors vs "blank slate" minds) have been going on for many decades, across multiple fields, and deep learners have *zero* knowledge of this context
-
-
-
The key issue here is that AI related majors have been booming since 2014, by a large multiple, and they've been largely primed to believe "[SGD] is our answer, you don't need to know about anything else". That's intellectual blinkers, not an education
Show this thread
End of conversation
New conversation -
-
-
This Tweet is unavailable.
-
It is not true that the function approximation methods pre-1940 are the same ones that deep learning is using today. For example, Tikhonov's book on regularization was published in 1977
- Show replies
-
-
-
What are some books/online material you think would provide interesting background knowledge?
-
Aristotle. Physics.
- Show replies
New conversation -
-
-
agreed. i've got a bio-med background with a few years of psychology and i think a lot about innate priors as i'm yelling at my RL ragdolls to stop trying to use their face as a foot.
-
I originally read that as “IRL ragdolls” and thought “damn, that’s a dark way to describe your children”
End of conversation
New conversation -
-
-
i've been working on reward functions that are able to adjust themselves but even that seems like too much babysitting.
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
This Tweet is unavailable.
-
This Tweet is unavailable.
- Show replies
-
Loading seems to be taking a while.
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.