611–620 of 876 posts
Earlier quoted context omitted.
The techniques highlighted in this prize are not really that useful for deep learning.
You mean besides bringing it into existence at all?
Here is the reasoning: https://www.nobelprize.org/uploads/2024/09/advanced-physicsp... I'm surprised Terry Sejnowski isn't included, considering it seems to be for Hopfield Nets and Boltzmann machines, where Terry played a large role in the latter.
I find the prize a bit odd this time since it focused on Hopfield networks and Boltzmann machines. Picking those two architectures in particular seems a bit arbitrary. Besides, Parisi got the prize last year (edit: actually 2021, time flies) for spin glasses. Hopfield networks are quite related. They could have included Hopfield & Hinton too, and it would have looked more coherent. It is also concerning that lately t…
Parisi won in 2021, not last year. His work was more about establishing spin glasses as a way to study complex systems. Hopfield definitely built on that, showing how those ideas could be applied to neural networks and info storage in state-space machines. As for focusing on Hopfield networks and Boltzmann machines, I get where you're coming from. They’re just a couple of architectures among many, but they’re pretty…
I definitely think their work is deserving of awards, but I kinda agree with other commenters in that this says more about the Nobel committee than anything i.e. Hinton has already won a Turing Award in 2018, and there is no Nobel for computer science And this work was already recognized to have impact ~12 years ago, when he auctioned his company of 2 grad students to Google/Microsoft/Baidu/Facebook, for over $40M, u…
>scientific awards are more interesting when leading indicators
Peter Higgs waited 50 years, the Nobel is not a "leading indicator." If it was, it would be given out on the basis of the "hype cycle," which would not be very helpful to anybody.
Earlier quoted context omitted.
Geoffrey Hinton wins the Nobel prize in physics for giving physics postdocs more reasonable job market options
This prize is more of a settler colonialist land grab by physicists. ML is just a subfield of physics (like every other field), so let's make sure that everyone knows that it's in our domain. Speaking as a onetime physicist now in ML...
Statisticians would never have done that due to parsimony and something something Bayesian.
Engineers would never have done it, nor mathematicians either.
It took Computer Scientists because it is computation.
Earlier quoted context omitted.
the fields medal fills that gap
Yeah, and the Turing Award for CS, and there's probably a bunch more that I'm forgetting. Still kind of weird that the prestigious award that everyone has heard of doesn't have a mathematics category.
Has NN led to any fundamental breakthrough in physics research? I understand the impact from NN to scientific research, but I'm not aware of any big results in fundamental physics research because of NN.
Physics is the study of the physical world, and learning, imagination, creativity, are all phenomena that we observe in the physical world but have only a primitive understanding of. It's a staggering advancement that we can now simulate key aspects of each.
Earlier quoted context omitted.
The Swedish central bank later added the price in economics (although controversial). So other things can be amended. However, I suspect we won't see any more amendments to the noble price. A different institution can step up and make a price on its own, though I'm not sure what institution would have that amount of prestige without weaponizing it for commercial purposes. I prefer a nicer price for mathematics that i…
> I don't think there is much room for a "society-changing innovation" within CS that isn't either an engineering feat (Linux, Docker, FFmpeg) or an algorithm that could fit under a mathematics price (FFT, Navier Stokes). I'm not sure I agree with that. There's plenty of theoretical computer science that isn't really "engineering" and would fall into a pretty different category than stuff like FFT or Navier Stokes. I…
I would still consider CSP, Petri Nets, and Pi-Calculus mathematical enough to be wrapped under a mathematics price if they're influential enough. The first true computer scientists were mathematicians, and I still feel that much of the theoretical work in the field is closer to "mathematics useful for computers" than its separate field.
In the spirit of the nobel price, "mathematics with the greatest humanitarian impact" leaves plenty of room for the inclusion of influential pieces from theoretical computer science, especially as those prices within mathematics that do exist already include loads of mathematics that require computers to prove or solve.
Has NN led to any fundamental breakthrough in physics research? I understand the impact from NN to scientific research, but I'm not aware of any big results in fundamental physics research because of NN.
The simplest cases will have been long enabled by simpler regressions and such, of course, but the more complex pattern recognition appears to be appreciated.