[Labo333]

I'm not going to comment on the actual content that is mostly [1] scientifically correct, but Schmidhuber (the author) has a record of wanting to be the center of attention [2] (even though LeCun is not better on that matter). Also, a third of the sources are written by him...

Just look at his previous blog post [3], in which he explains that the most cited neural networks all cite works by him. These papers cite dozens of papers, so a lot of other groups that are active in AI can claim the same thing...

[1]: For example, Turing published an independent proof of the Entscheidungsproblem, in the [TUR] article, just a month after Church, that the article forgets to highlight.

[2]: https://en.wikipedia.org/wiki/J%C3%BCrgen_Schmidhuber#Views

[3]: https://people.idsia.ch/~juergen/most-cited-neural-nets.html

[albertzeyer]

He just wants to get the facts right, esp the correct attribution to the original scientific contributions (who did it first).

Originality is easily defined as who did sth first.

This might not be the same as influence of some work. It might be that someone else does a lot of groundbreaking work which actually makes sth work (e.g. Goodfellow et al for GAN). You can say the GAN paper had more influence than Schmidhubers Adversarial Curiosity Principle.

Also, of course some newer authors might not know of all the old work. So it might be that people get the same ideas. So when Goodfellow got the idea for GAN, he might not have known about Schmidhubers Adversarial Curiosity.

The problem is, sometimes people did know about the other original work but intentionally do not cite them. You can not really know. People of course will tell you they did not know. But this can be fixed by just adding the citation. It looks bad of course when there are signs that they should have known, so it was really intentionally.

There is also a lot of arguing when sth is the same idea, or when sth is a different novel idea. This can be ambiguous. But for most cases which are discussed by Schmidhuber, when you look at the core idea, this is actually not so much the case. Also, this is also not so much a problem. There is less argumentation about whether sth is at least related. So this still should be cited then.

The question is then, which work should one cite. I would say all the relevant references. Which is definitely the original work, but then also other influential work. Many people just do the latter. And this is one of the criticism by Schmidhuber, that people do not give enough credit (or no credit) to the original work.

[IshKebab]

> of wanting to be the center of attention

It seemed more like he felt he was unfairly being uncredited. Which is probably why he wrote this - he now cares deeply about giving credit to the right people.

[bodge5000]

Surely the more noble cause for that would be giving more credit to others, rather than attempting to take away credit from a well known figure. This article is somewhat about the other important figures who's knowledge Turing's was built off, but its central point is that Turing gets too much credit.

I understand why he'd care about that if he'd been uncredited and watched peers be overcredited, but I'd hardly call it a noble work, even if it is understandable.

[6gvONxR4sf7o]

The article is full of credit given to a huge number of people.

[bodge5000]

The article is called Turing oversold, and the article is all about who should be getting credit instead of Turing. This isn't "Hey, are you aware of all these people who helped develop computer science", its "Turing is overcredited, heres a list of other people to support my argument"

[IshKebab]

I disagree. It read more like "Turing is overrated, you should credit these people instead."

[YeGoblynQueenne]

>> I'm not going to comment on the actual content that is mostly [1] scientifically correct, but Schmidhuber (the author) has a record of wanting to be the center of attention [2] (even though LeCun is not better on that matter).

You_again wants his work and that of others properly recognised. For example, his article, titled Critique of Paper by "Deep Learning Conspiracy" (Nature 521 p 436) [1] that is referenced by your link to wikipedia, cites a couple dozen pioneers of deep learning, quite apart from Schmidhuber hismelf. Quoting from it:

>> 2. LBH discuss the importance and problems of gradient descent-based learning through backpropagation (BP), and cite their own papers on BP, plus a few others, but fail to mention BP's inventors. BP's continuous form was derived in the early 1960s (Bryson, 1961; Kelley, 1960; Bryson and Ho, 1969). Dreyfus (1962) published the elegant derivation of BP based on the chain rule only. BP's modern efficient version for discrete sparse networks (including FORTRAN code) was published by Linnainmaa (1970). Dreyfus (1973) used BP to change weights of controllers in proportion to such gradients. By 1980, automatic differentiation could derive BP for any differentiable graph (Speelpenning, 1980). Werbos (1982) published the first application of BP to NNs, extending thoughts in his 1974 thesis (cited by LBH), which did not have Linnainmaa's (1970) modern, efficient form of BP. BP for NNs on computers 10,000 times faster per Dollar than those of the 1960s can yield useful internal representations, as shown by Rumelhart et al. (1986), who also did not cite BP's inventors.

That is not "wanting to be the center of attention". It is very much asking for proper attribution of research results. Failing to do so is a scientific scandal, especially when the work cited has contributed towards a Turing award.

__________

[1] https://people.idsia.ch/~juergen/deep-learning-conspiracy.ht...