Hacker News Reader

Hacker News

Comment by daniel-thompson | Hacker News Reader

Parent Full thread

daniel-thompson

Can you elaborate on this.

View on HN

Let's just say that is it correct that he is still known for his 2014 GAN work.

Anonymously trashing someone on the internet like this is pathetic.

It must be difficult to repeat the success of a once-in-a-lifetime finding.

I didn't take it to mean that, exactly. GANs are a very impressive trick. But their social cachet is not, as far as I'm aware, in proportion to people's success at using them for practical applications that generate revenue. So the implication might be, "Don't expect a brain drain in Apple's GAN department to have a significant impact on their business fortunes."

You think GANs are once in a lifetime?

Given its impact on ML, for a single person to discover, yes.

he's also known for single handedly starting the field of adversarial examples, which is a huge field. he basically started two fields. he's huge

GANs were just low hanging fruit that Schmidhuber already found. By most accounts Ian Goodfellow is kind of an asshole, and there was some drama about him a few times over the years. I don’t know the man, that’s just what I’ve heard.

If Schmidhuber had really found everything he says he found, the Singularity would've happened decades ago.

You know, if it were anyone else I would probably ignore this comment, but you seem to do your homework, to put it mildly. Are you saying he's, uh, misrepresenting himself?

(I like Schmidhuber but I'm not knowledgeable enough to really evaluate his work or claims.)

he's had a decades-long tendency to just say that stuff has been invented before, sometimes by him, sometimes by obscure russians, that one time by Gauss. sometimes he's right, more of the time he's materially wrong, all of the time he's got no social sense.

To be fair, it’s not just him. Whenever I gave a poster at a conference, there was a chance that someone with a Russian sounding name would call the work trivial and say they had already published it in the nineties ;). Of course, on closer inspection that was not the case.

I don’t fully understand the tendency to do so, but it appears to be fairly widespread.

Is there anything that schmidhuber hasn't done before everyone else?

Remote work.

So he was a good culture fit for apple then?

"Director of Machine Learning" is different than "A director of a machine learning project." "Director of Machine Learning" kind of implies that he oversaw a lot of machine learning projects across the company. Being "a director of a machine learning project" means you run a team. Big difference.

I guess citations do not directly translate into products?

No, especially in fields that have as much citations and papers on average as ML/DL/RL.

If a paper in any other field was cited 44k in ~8 years, it must have made life-changing discoveries. Maybe some of the early papers on COVID-19 will reach that number.

The implication I thought was that his credits might have ended in 2014

I assume he just thinks Siri sucks?

Reply on news.ycombinator.com