I think that “rational, deliberate design”, as you put it, is simply far less common (than random chance) than you think; that the vast majority of human knowledge is a result of induction instead of deduction; that theory is overrated and experimentalism is underrated.
This is also why I highly doubt that anything but prosaic AI alignment will happen.
Yeah. Here’s an excerpt from Antifragile by Taleb:
One can make a list of medications that came Black Swan–style from serendipity and compare it to the list of medications that came from design. I was about to embark on such a list until I realized that the notable exceptions, that is, drugs that were discovered in a teleological manner, are too few—mostly AZT, AIDS drugs.
It is likely that no new materials were discovered. Most of the materials produced were misidentified, and the rest were already known. This happened because of problems in both the computational and the experimental portions of the work.
The computational predictions suffered as they generated structures with ordered cations, where in fact the same or very similar compounds are known with disordered cations. This inability to deal with compositional disorder is an important limitation of the methods used here. When the predicted, ordered, compounds were synthesised experimentally, the known, disordered, compounds were produced instead. This accounted for 2⁄3 of the whole list of compounds in the paper.
(...)
My personal view hasn’t changed from my initial analysis. This paper is incorrect in its headline claims. No new materials were discovered. The best thing to happen here is for the paper to be retracted while these fundamental issues are fixed, both on the computational and experimental side.
Does this count as “rational, deliberate design”? I think a case could be made for both yes and no, but I lean towards no. Humans who have studied a certain subject often develop a good intuition for what will work and what won’t and I think deep learning captures that; you can get right answers at an acceptable rate without knowing why. This is not quite rational deliberation based on theory.
But it shows that you don’t necessarily need to rely strictly on experimentation! Certainly it still relies on it, but humans have been doing this sort of thing for a while. While I agree it’s the case that people still have to do a lot of experiments historically, it’s quite possible to have very detailed sketches of what can and can’t be done.
Doesn’t have any bearing historically. Also seems more like a brute force search, where the component of studying the materials properties has been made more efficient (by partially replacing lab experiments with deep learning).
The entire purpose of the paper is eliminating a brute force search. It wouldn’t be possible to identify these materials with brute force. Deep learning lets you bypass brute force by zooming in on the shapes in the energy landscape that are relevant to what you’re doing. A similar thing is possible with human intuition. It’s certainly not the case that this has allowed people to completely avoid experimentation, but I share it to point out that it’s not actually impossible to have very strong models of the energy landscape.
I think that “rational, deliberate design”, as you put it, is simply far less common (than random chance) than you think; that the vast majority of human knowledge is a result of induction instead of deduction; that theory is overrated and experimentalism is underrated.
This is also why I highly doubt that anything but prosaic AI alignment will happen.
Yeah. Here’s an excerpt from Antifragile by Taleb:
https://deepmind.google/discover/blog/millions-of-new-materials-discovered-with-deep-learning/
This was apparently a bust:
Does this count as “rational, deliberate design”? I think a case could be made for both yes and no, but I lean towards no. Humans who have studied a certain subject often develop a good intuition for what will work and what won’t and I think deep learning captures that; you can get right answers at an acceptable rate without knowing why. This is not quite rational deliberation based on theory.
But it shows that you don’t necessarily need to rely strictly on experimentation! Certainly it still relies on it, but humans have been doing this sort of thing for a while. While I agree it’s the case that people still have to do a lot of experiments historically, it’s quite possible to have very detailed sketches of what can and can’t be done.
Doesn’t have any bearing historically. Also seems more like a brute force search, where the component of studying the materials properties has been made more efficient (by partially replacing lab experiments with deep learning).
The entire purpose of the paper is eliminating a brute force search. It wouldn’t be possible to identify these materials with brute force. Deep learning lets you bypass brute force by zooming in on the shapes in the energy landscape that are relevant to what you’re doing. A similar thing is possible with human intuition. It’s certainly not the case that this has allowed people to completely avoid experimentation, but I share it to point out that it’s not actually impossible to have very strong models of the energy landscape.
Sorry, low effort comment on my side. Still, I think the original link seems misleading in the point it’s purportedly trying to make.