Hacker News Reader

Comment by pishpash | Hacker News Reader

They're just titles. The patentable claims are obviously more narrowly constrained.

"Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for environment simulation. In one aspect, a system comprises a recurrent neural network configured to, at each of a plurality of time steps, receive a preceding action for a preceding time step, update a preceding initial hidden state of the recurrent neural network from the preceding time step using the preceding action, update a preceding cell state of the recurrent neural network from the preceding time step using at least the initial hidden state for the time step, and determine a final hidden state for the time step using the cell state for the time step. The system further comprises a decoder neural network configured to receive the final hidden state for the time step and process the final hidden state to generate a predicted observation characterizing a predicted state of the environment at the time step."

From "Recurrent Neural Networks" patent. This doesn't sound narrowly constrained to me. This just sounds like a RNN, which according to wikipedia, were conceived in 1982.

https://en.wikipedia.org/wiki/Recurrent_neural_network

That is the abstract of an application--it does not define the protection of a patent. It is intended to aide someone browsing patents to determine if this document may be something worth looking into closer, it is supposed to be quite broad and vague.

To add, here is more detail: https://worldwide.espacenet.com/publicationDetails/claims?CC...

https://register.epo.org/documentView?number=US.2017058046.W...

Read the applications for yourself. I'm not a patent lawyer but they seem to be unbelievably broad.

That is what I see, as well, with the GENERATING AUDIO USING NEURAL NETWORKS patent application. IANAPL, but I do work in audio software and the application, seems to me, explains a broad model of audio processing using convolutional neural networks (in the claims an input is involved, so I am guessing the title tried to select an even broader set of problems - with & without input).

I think people pointing out that Google defensively files patents may be correct and I hope that's the intention with DeepMind's work, too. I am a fan of the audio work they did with WaveNet since day 1 and was so happy the paper was out for all of us to read. Still, I was hoping I may be able to (read: "daydreaming that I might be capable to") build on it, some day.

Organizations change leadership and they switch focus, too. I believe patents slow-down (even stop) progress and they are quite hypocritical, overall (show me a patented invention that was so unique, the author/owner didn't "stand on the shoulders of" 100's of years of freely available science to derive/discover/invent it and... I will agree for that specific case).

pishpash is right folks, put down your pitchforks until claims are granted. Then decide if you want to be outraged.

Yes, let's not get angry until after they successfully monopolize AI algorithms that they never created.

No, really, you shouldn't. We do not know what they are going to 'monopolize' yet. A patent application, even if published, is not enforceable, it is not a patent. I could file an application today claiming to have invented the wheel, and 18 months from now it would publish as an official looking patent publication. It means literally nothing.

Can't we be mad at DeepMind for filing them? Even if in the end they are not granted?

Because of the way US patents work, it really is a "We should file these so somebody else doesn't claim-jump them and put us in the position of defensively proving prior art" system.

As a deep-pocket company, Google can afford to file patents and has a track record of being responsible about them in the community. It makes about as much sense to be mad at DeepMind for not filing them.

Reply on news.ycombinator.com