caenorst

I have spent enough time to debunk papers to know that a lot of researchers are shamelessly cheating or make mistakes (as EVERYBODY does). Reviewing your code should be part of the peer review process.

Sorry but I don't trust your research just based on yours words and cherry picked images.

Also research is incremental, so producing a proper code on which other can work on top of should be part of the CONTRIBUTION.

The angle of approach is different. Steganography is about hiding a message, you want it to be as undetectable as possible (including using technology). Eventually a transformation may completely destroy the hidden message, but you are supposed to transmit the media as is. The usage is mostly for spying or crime.

Digital Watermarking, you want the "message" to be very robust to transformation. The main usage I know is to help protection against piracy, for example if somebody upload a movie online, you want to put a specific signature related to the location where you sold the movie so you can find the uploader. Eventually the uploader will try to remove the watermarking by either identifying it or by applying some transformation to the movie in hope that it will make the message unrecognizable. So it's a game between industry and the potential tricks that the uploader could apply.

The intersection between the two fields is that they both try to modify the media as little as possible.

Steganography is about hiding not robustness.

The point of view of this work is much more related to watermarking than steganography, as they don't really focus on detectability.

Industries have been using watermarking for a long time, for example to identify video piracy.

I'm not saying that there's no very big model, just saying that it's a minority of publications, for any trivia example of big model I can show you 10x trivia examples of relevant non-big models.

Also you are talking about a model which is specifically designed for TPU (the dimensionality of the networks is especially fine-tuned).

And even tho, BERT_large still fit in the memory of a single GPU (for very small batch), there is an implementation on Pytorch. I don't understand, are people complaining that Deep Learning is actually (reasonably) scaling ? Isn't it a good news ?

Sorry but even tho the big companies produce a lot of interesting research, I challenge you to not find any interesting model trained on a single GPU from recent publications (the majority coming from academia). Actually it's very rare to find a paper where largely distributed training is necessary (i.e: the training would fail or would be unreasonably too long). Yes having more money help you to scale your experiments, it's nothing new and it's not something specific to AI.

Unfortunately most of people out of France don't understand the situation here. They get to blame some violence propaganda of twitter account with less than a hundred followers without talking about how one-sided are mainstream medias (which have much bigger reach). I'm tempted to say that this is a move to discredit Gilets Jaunes movement.

That's not even an argument. Here is one against the claims.

Modification: There have been repeatedly successful case of transfer learning where you take a pretrained deep neural network and train on a new dataset (even very small).

Vulnerability: you dont need the training dataset to test vulnerabilities, you just either very specific input (that you probably gonna have to create yourself because they are so specific that they are not even in the dataset) or using one input that you corrupt for attacking. And as far as I know it doesn't sound harder than making test for regular algorithms.

What makes you say that you can't modify a model ? And can't inspect it for vulnerabilities ?

Summarizing the social liberal aspect of Canada by the legalization of marijuana seems very narrow minded. There are far more important social aspects than marijuana legalization.

Did they communicate any release date and price during the show ?