Hacker News Reader

Comment by jowday | Hacker News Reader

Any claim Schmidhuber might have is always drowned out by how obstinate he is about demanding recognition.

He’s interrupted lectures at major conferences from other leaders in in the field to quibble about citations unrelated to the subject of the lecture - pretty much everything he does involves some whining about how the credit for his work has been stolen from him. It’s hard to take his claims seriously when he’s this pigheaded.

View on HN

Miron Livny does this too. He created the Condor project (a batch queue/executor system) and always interrupted talks about distributed computing to say he did it first.

Very few peoople used Condor because it was too hard to configure.

There’s a blast from the past… I actually built a headless opportunistic cluster using condor back in grad school so I could steal cpu cycles from the windows computer lab over nights and weekends.

https://www.linux.com/news/condor-building-linux-cluster-bud...

For what it's worth, even in that space, NQS was done before Condor (and based on an earlier system).

I've seen Condor used quite heavily around for long time, but originally only PVM, which was roughly contemporaneous.

yes but condor added a number of things that other folks later copied and (according to livny) didn't provide enough credit for, including (but not limited to) containers, glide-in, federation, secure communications, matchmaking (classads).

I actually like Schmidhuber, and I don't mind his boisterous arrogance, but you have to admit there was a heavy dose of Schadenfreude there in him missing out of the Turing awarded jointly to Hinton, LeCun, and Bengio.

This is the one bit where I felt he was under-recognised. LSTMs for example are 100% his team’s work and that’s been a major thread in ANNs. Plus all the other bits...

Yeah, obnoxious or not, the Turing award is a bit too important for this kind of middle school drama.

Can you provide more information about him?

The mental image I got purely from reading the blog/paper is quite intriguing...

His schtick is claiming he’s real “father of deep learning” and that all major work in ML has been built off of work done in his lab, and that he’s unjustly remained unrecognized while other pioneers in the field have received tons of awards and compensation from big tech companies.

When I say it’s his shtick, I mean that he basically brings it up whenever possible. When another big name in the field receives an award, he’ll write up a huge blog post about how ackshually they didn’t invent XYZ, but rather modified A and B, which is related to work C he did at his lab in 1980.

See https://people.idsia.ch//~juergen/critique-honda-prize-hinto...

https://people.idsia.ch//~juergen/naturedeepmind.html

https://people.idsia.ch//~juergen/critique-turing-award-beng...

At 1:03:00 in this video he interrupts a lecture at NIPS from Goodfellow (another leader in the field) to try and claim credit for the subject of the lecture in a sort of slimy way. https://youtube.com/watch?v=HGYYEUSm-0Q&feature=share

Honestly I’m not equipped to tell whether Schmidhuber’s claims are true since I haven’t followed the progression of ML research since the 1980s - but the impression I get of him is that he’s kind of a jackass who’s willing to spend a ton of time tearing others down in really petty ways. If there is some truth to his claims, perhaps the reason he’s unrecognized is because he’s a bit of an ass to work with?

I met him, he’s very nice in person (at least over the course of a day, say). Kind, patient and genuinely interested in other people.

So I doubt it is as simple as he’s a horrible person. But clearly he’s an outlier in terms of demanding recognition.

Tangent: While I've listened to him on podcasts, and am reading his Deep Learning textbook, this is the first time I've seen Goodfellow in person. I'm crushed to see that he is not, as I assumed, some 60 year old professor who has spent decades earning his expertise, and instead is someone young enough to pass as a college student!

A lot of context in his own post here, and the reply from Dr. Hinton is linked:

https://people.idsia.ch//~juergen/critique-honda-prize-hinto...

I stopped reading at [backpropagation] "wasn't created by "lots of different people" but by exactly one person who published first [BP1] and therefore should get the credit". Okay, so he was the first. Brilliant insight. Hugely valuable. He should get credit for that, and I'm very sympathetic if that hasn't happened (I don't follow other research enough to make any strong statement about if he has or hasn't received appropriate recognition). But that was 50 years ago and many equally insightful and impactful developments have gotten us to the point where backpropagation is being done in a practical way. "Exactly one person" is ridiculous.

I saw him speak like 6-7 years ago and it was pretty clear that he was trolling and didn't take himself seriously.

Reply on news.ycombinator.com