Thanks, I know the article. I thought it was interesting to see that you-again is not alone in attributing the progress of deep learning to Japanese and Finnish researchers who lack wider recognition.
I think a good analogue for this situation is that of the Nobel in Physics for the Blue LED.
The first blue LED was built by Herb Maruska and Wally Rhines at Stanford in 1972. However it was incredibly dim.
It took decades of focused and sustained research by Shuji Nakamura and others to make a commercially viable blue LED. Hence Nakamura won the Nobel
Same deal with AI/ML.
Although Nakamura was actually doing something PHYSICS related!
If we go back in time the perceptron was invented in the 40s by Warren McCulloch and Walter Pitts. The first hardware implementation was Mark I Perceptron machine built in the 50s [1].
“We must not forget the role played by pioneer Japanese researchers in erecting the foundations of neural network research.”
By awarding the Nobel to Hopfield and Hinton a lot of folks were "forgotten" but that's not the right way to look at a Nobel. You can't award everyone that ever contributed to such a large innovation through the century. It's about recognition and not forgetting, even when a narrow selection has to be made. Just enjoy the fact that your beautiful field was awarded such a prize and recognition it truly deserves.
Is back propagation really an idea applicable to the Nobel prize in physics? I mean certainly a Turing award is warranted. But back propagation or neural networks don’t really tell us anything about how the universe works. It’s an efficient optimization algorithm for a really powerful class of models.
The 2024 Physics Prize is not being awarded for backpropagation but “for foundational discoveries and inventions that enable machine learning with artificial neural networks”.
They seem to be focusing on energy function, statistical mechanics connection, etc. of Hopfield Nets and Boltzmann machines.
An auxiliary motivation seems to be the use of ANN:s becoming a significant tool for very-big-data physics.
This is a good article. My impression is the way we think of scientific credit in general is broken. There's too much of an attempt to attribute credit to a small number of people. This might or might not have been more accurate at one time in history but it's increasingly a problematic schema for how to think about how science works.
I don't mean that people who get credit don't deserve it, more so that there are often a series of people who also deserve it and don't get credit.
21 comments
[ 9.4 ms ] story [ 45.9 ms ] threadhttps://news.ycombinator.com/item?id=42264069
If dang is watching I would like recommend the earlier article for the repost queue.
Annotated History of Modern AI and Deep Learning - https://people.idsia.ch/~juergen/deep-learning-history.html
Pretty good and mentions a lot of folks i didn't know about.
The first blue LED was built by Herb Maruska and Wally Rhines at Stanford in 1972. However it was incredibly dim. It took decades of focused and sustained research by Shuji Nakamura and others to make a commercially viable blue LED. Hence Nakamura won the Nobel
Same deal with AI/ML.
Although Nakamura was actually doing something PHYSICS related!
[1] https://en.wikipedia.org/wiki/Perceptron
Science is a large network of contributions and collaboration.
They seem to be focusing on energy function, statistical mechanics connection, etc. of Hopfield Nets and Boltzmann machines. An auxiliary motivation seems to be the use of ANN:s becoming a significant tool for very-big-data physics.
Nobel Foundation descriptions (from discussion at the announcement https://news.ycombinator.com/item?id=41775463 ):
https://www.nobelprize.org/uploads/2024/09/advanced-physicsp...
https://www.nobelprize.org/prizes/physics/2024/summary/
You are from Sweden or Finland, right?
I don't mean that people who get credit don't deserve it, more so that there are often a series of people who also deserve it and don't get credit.
This is normal. See "The lifecycle of a troubled project". The last step is "Dictinctions for the uninvolved".
"There's a tendency among the press to attribute the creation of a game to a single person," says Warren Spector, creator of Thief and Deus Ex.