Roger Critchlow points us to this post from Jürgen Schmidhuber, “How 3 Turing Awardees Republished Key Methods and Ideas Whose Creators They Failed to Credit.”
The whole thing is too long and detailed to follow—it’s like one of those pieces of outsider art with scribblings all over the margins of an elaborate painting on an already-patterned fabric—but I did follow one of the links in the post which pointed to this:
This reminds me of Stan’s pedantic mode where it gives a warning if you do something like define a variable named sigma without constraining it to be positive or if you assign an inverse-gamma(0.001, 0.001) prior distribution. Ask Facebook’s AI about Jürgen Schmidhuber and you get a warning!
I have no idea what to think about the priority dispute. This sort of thing comes up in statistics; see for example this discussion of attribution for the potential outcomes model in causal inference, summarized by this comment from Guido Imbens.
Critchlow also points to this thread on Hacker News which has lots of takes in all directions.
My proposed solution to the problem
Jürgen Schmidhuber works at the King Abdullah University; Geoffrey Hinton was educated at Kings College; Yoshua Bengio is a Fellow of the Royal Society; Yann LeCun is a Chevalier of the French Legion of Honour; and Bengio, Hinton, and LeCun have received the Princess of Asturias Award.
That’s two kings, a royal, a knight, and three princesses.
Can’t they just settle this like aristocrats?
Do you mean via duels?
Or he means trace it back to Charlemagne? Then everyone has equal priority:
https://en.wikipedia.org/wiki/Royal_descent
I thought that, too! Which begs the question, swords or guns?
On a more serious note, I cannot think of King Abdullah University without thinking of bought authorships (https://english.elpais.com/science-tech/2023-04-18/saudi-arabia-pays-spanish-scientists-to-pump-up-global-university-rankings.html#).
I am currently reading James Gleick’s book entitled “Isaac Newton,” and pages and pages are filled with Newton’s disagreements and rages against others. To see that stuff has not changed that much since then,
https://en.wikipedia.org/wiki/J%C3%BCrgen_Schmidhuber
“The term “schmidhubered” has been jokingly used in the AI community to describe Schmidhuber’s habit of publicly challenging the originality of other researchers’ work, a practice seen by some in the AI community as a “rite of passage” for young researchers. Some suggest that Schmidhuber’s significant accomplishments have been underappreciated due to his confrontational personality.”
Llama 3 (I tried it with the model that https://gpt4all.io/ uses by default) is much nicer to him:
Jürgen Schmidhuber is a German computer scientist and professor who has made significant contributions to the field of artificial intelligence (AI) and machine learning. He is particularly known for his work on deep learning, recurrent neural networks (RNNs), and long short-term memory (LSTM) networks.
Schmidhuber received his Ph.D. in 1996 from the University of Stuttgart and has since held positions at various institutions, including the Technical University of Munich (TUM) and the German Research Center for Artificial Intelligence (DFKI). He is currently a professor at TUM’s Department of Computer Science and Engineering.
Schmidhuber’s research focuses on developing new AI algorithms that can learn complex patterns in data. His work has led to several breakthroughs, including:
1. Recurrent Neural Networks (RNNs): Schmidhuber developed the first RNN architectures capable of learning long-term dependencies.
2. Long Short-Term Memory (LSTM) networks: He introduced LSTMs as a solution for vanishing gradients in traditional RNNs and has since worked on improving their performance.
3. Deep Learning: Schmidhuber’s work on deep neural networks has led to state-of-the-art results in various AI applications, such as image recognition, speech recognition, and natural language processing.
Schmidhuber is also known for his enthusiasm for promoting AI research and education. He founded the German Research Center for Artificial Intelligence (DFKI) and serves as its director. Additionally, he has written several books on AI and machine learning, including “Deep Learning” and “Recurrent Neural Networks”.
Overall, Jürgen Schmidhuber is a renowned expert in the field of artificial intelligence and machine learning, with significant contributions to the development of deep learning algorithms and their applications.
Carlos Ungil: Does a different AI site echo the Schmidhuber enthusiasm of the above? If not, why not? If so, does that settle the issue?
I don’t know what issue needs to be settled. I was just pointing out that Facebook’s AI – as Andrew called it – seems to be less confrontational at 3 than at 2. It may be growing up.
> Yann LeCun is a Chevalier of the French Legion of Honour
> That’s two kings, a royal, a knight, and three princesses.
Interestingly the “Légion d’honneur” is a republican institution created by Napoleon and the first grade – originally “légionnaire” – became “chevalier” in a subsequent monarchy period. They also added “Royal” to the name of the order and while that adjective was dropped later the “legionary” title was not reinstated.
Funny I thought of a decidedly more off color reference
https://en.m.wikipedia.org/wiki/The_Aristocrats