Live data from Hacker News

LeCun's 2022 paper on autonomous machine intelligence does not cite prior work

people.idsia.ch

21–30 of 95 posts

Re: LeCun's 2022 paper on autonomous machine intelligence does not cite prior work

#21
post #15

background from the great Genius Makers history of ML: The press needed heroes for its AI narrative. It chose Hinton, LeCun, Bengio, and sometimes Ng, thanks largely to the promotional efforts of Google and Facebook. The narrative did not extend to Jürgen Schmidhuber, the German researcher based on Lake Lugano who carried the torch for neural networks in Europe during the 1990s and 2000s. Some took issue with Schmidh…

Reading some Schmidhuber's writings on this issue is a great way to learn the history and development of AI and ANN's that is often missed. At the same time Schmidhuber's tone and anal-retentive attitude explains why he is considered so annoying. The field is full of huge egos and overly simplified narratives from a singular point of view. For example, it's true that Hinton was a torchbearer of neural networks for se…

I believe it's reasonable for him to request a citation if significant results of his works are quoted.

LeCun would probably also be upset if he read about his work in the newspaper, but mis-attributed to someone else.

Also, millions in funding money depend on stuff like this. If LeCun can pretend to be the inventor, he gets the grant money. If Schmidhuber gets the citation he seeks, he can use that to apply for new funding. So it's not just prestige, these citation issues have a large monetary impact, too.

Re: LeCun's 2022 paper on autonomous machine intelligence does not cite prior work

#22
post #11

Earlier quoted context omitted.

We are scientists and can use names. 'him' is Jürgen Schmidhuber. He's a luminary of ML, but he has entered into several attribution controversies. He famously accused Ian Goodfellow of appropriating his work to create generative adversarial networks (GANs) in a public venue (a NeurIPS workshop). I agree with the parent comment. Ultimately, Science is about advancing the human condition more than celebrating great in…

>Ultimately, Science is about advancing the human condition more than celebrating great individuals. Giving proper credit to prior work is a fundamental component of scientific progress. Just look at the Chinese academic system to see what happens to science when people are given free reign to plagiarise.

> Just look at the Chinese academic system to see what happens to science when people are given free reign to plagiarise.

Can you explain what this means? It's a pretty broad statement.

Re: LeCun's 2022 paper on autonomous machine intelligence does not cite prior work

#23

background from the great Genius Makers history of ML: The press needed heroes for its AI narrative. It chose Hinton, LeCun, Bengio, and sometimes Ng, thanks largely to the promotional efforts of Google and Facebook. The narrative did not extend to Jürgen Schmidhuber, the German researcher based on Lake Lugano who carried the torch for neural networks in Europe during the 1990s and 2000s. Some took issue with Schmidh…

In my opinion the difference that Hinton students made was using GPUs to run deep learning models. First in 2010, when publishing Theano at NIPS, making deep learning on a single computer node with a large GPUs very approachable ( using a Python framework, precursor to TensorFlow). Then in 2012, with Alex Krizhevsky masterfully using two GPUs to build the landmark model AlexNet, which outperformed traditional compute…

Funny ...

According to [1], the papers [42,43] showed that plain backprop was capable of solving complex tasks without pretraining on a GPU.

From [1]:

> Our team then showed that good old backpropagation [A1] on GPUs (with training pattern distortions [42,43] but without any unsupervised pre-training)

I can't verify since I have a barely stable connection and little battery, but somebody else should be able to.

[1] https://people.idsia.ch/~juergen/firstdeeplearner.html

[42] H. Baird. Document image defect models. IAPR Workshop, Syntactic & Structural Pattern Recognition, p 38-46, 1990

[43] P. Y. Simard, D. Steinkraus, J.C. Platt. Best Practices for Convolutional Neural Networks Applied to Visual Document Analysis. ICDAR 2003, p 958-962, 2003.

Re: LeCun's 2022 paper on autonomous machine intelligence does not cite prior work

#24
post #12

Lecun's post is not even a paper as stated in the foreward of the paper. But thanks for bringing it up, i will read it > This document is not a technical nor scholarly paper in the traditional sense, but a position paper expressing my vision for a path towards intelligent machines that learn more like animals and humans, that can reason and plan, and whose behavior is driven by intrinsic objectives, rather than by ha…

"intrinsic objectives" and "hard-wired programs" sound alike. Given that both must be designed.

Re: LeCun's 2022 paper on autonomous machine intelligence does not cite prior work

#25
post #24
post #12

Lecun's post is not even a paper as stated in the foreward of the paper. But thanks for bringing it up, i will read it > This document is not a technical nor scholarly paper in the traditional sense, but a position paper expressing my vision for a path towards intelligent machines that learn more like animals and humans, that can reason and plan, and whose behavior is driven by intrinsic objectives, rather than by ha…

"intrinsic objectives" and "hard-wired programs" sound alike. Given that both must be designed.

Meaning with defined goals (intrinsic objectives) instead of defined operations (hard-wired programs).

Re: LeCun's 2022 paper on autonomous machine intelligence does not cite prior work

#26
post #11

Earlier quoted context omitted.

We are scientists and can use names. 'him' is Jürgen Schmidhuber. He's a luminary of ML, but he has entered into several attribution controversies. He famously accused Ian Goodfellow of appropriating his work to create generative adversarial networks (GANs) in a public venue (a NeurIPS workshop). I agree with the parent comment. Ultimately, Science is about advancing the human condition more than celebrating great in…

>Ultimately, Science is about advancing the human condition more than celebrating great individuals. Giving proper credit to prior work is a fundamental component of scientific progress. Just look at the Chinese academic system to see what happens to science when people are given free reign to plagiarise.

Yes, like Watson & Crick and lots of others have shown over the years by lighting the way with proper accreditation. /s

Science is a means to an end, the 'giving proper credit' bit has been under fire since the day that someone figured out that citations can bestow fame (hardly every fortune) and given the low stakes it is to be expected that that is the battleground. But the number of PhD students that have been screwed out of proper credit to further aggrandize their academic masters can probably not be enumerated any more, so let's not pretend that proper credit is the coin that makes academia function.

Re: LeCun's 2022 paper on autonomous machine intelligence does not cite prior work

#27

Earlier quoted context omitted.

Well said. Attribution is important but it's impossible to know all the literature that is out there and even when you've tried your best to seek out prior arts there is always a possibility that you've missed something or not understood the connection with something previously read. Heck even when something is similar you may not cite it since it's so old the relevance is no longer there for a working research paper…

>Attribution is important but it's impossible to know all the literature that is out there and even when you've tried your best to seek out prior arts there is always a possibility that you've missed something or not understood the connection with something previously read. It's not like Schmidhuber's papers are obscure. >I've always thought that Schmidhuber was toxically uncharitable - what is the point of accusing…

> people stealing his ideas

So, ideas can now be stolen after all? I thought the HN consensus was that ideas can't be owned, but implementations can be. Or is it somehow different when we're looking at science?

Re: LeCun's 2022 paper on autonomous machine intelligence does not cite prior work

#29
post #24

Earlier quoted context omitted.

"intrinsic objectives" and "hard-wired programs" sound alike. Given that both must be designed.

Meaning with defined goals (intrinsic objectives) instead of defined operations (hard-wired programs).

A defined goal is just a defined operation with more steps.

It's inescapable.

Re: LeCun's 2022 paper on autonomous machine intelligence does not cite prior work

#30
There are 7 pages of citations. Not every somewhat related thought anyone has ever had or published will be cited. But incidentally, it is nerdily interesting to think of the bibliography section as a place to display contempt by omission that is subtle to most but glaring to few.
Post reply on HN