Hiya, I work at OpenAI. I think the Grover paper is a good place to read about some of this:https://arxiv.org/abs/1905.12616
We're likely publishing more on detecting fine-tuned outputs in the future, also.
Hi there, I work at OpenAI. This isn't related to OpenAI in any official sense, though there are some people involved who either have had links to us in the past, or currently have links to us.
I write a weekly AI newsletter called Import AI which is also cross-published to this WP blog. I try to cover a mixture of fundamental research papers and applied stuff. It also includes some OpenAI updates: https://jack-clark.net/
Hiya, (I'm Jack Clark) - this is a good point. The child probably gets about 50 to several hundred distinct 'frames' of the chicken. Still, a remarkably small number of examples.
There was an interesting academic research paper that showed you could train in GTA and transfer over to the KITTI dataset and do ok: https://arxiv.org/abs/1610.01983
It just instituted a bunch of changes which will radically change the magazine and may make it break even. "A person familiar with the matter said the publication, which Bloomberg acquired in 2009, was on pace to lose between $20 million and $30 million this year." http://www.wsj.com/articles/bloomberg-changes-businessweek-l...
Even more interesting - an OpenAI paper on the roughly same subject/technique came out a little earlier (RL2 fast reinforcement learning for slow reinforcement learning https://arxiv.org/pdf/1611.02779.pdf). I think parallel inventions tends to indicate that multiple people have stumbled on a similar good idea at the same time. (disclaimer: I work at OpenAI. v pleased to see these two papers emerge so close together)
Hiya, I wrote the article. What might not be captured here is that this is above all a community of people all working on AI so it's not like anyone is really emotional and/or aggressive about this issue. It's more that it's always better for an ecosystem to have multiple dominant players who are all relatively equal. Torch & Theano people were saying it's non-optimal if TF becomes the majority framework, which is a fair point. Everyone I spoke to for this article said TF was engineered well, albeit a bit of a black box in certain aspects.
Hi there, I wrote the article. We also looked at data on things like number of forks and contributors on GitHub as well as massive rise in questions on StackOverflow (thanks to Delip Rao's great post here http://deliprao.com/archives/168 and invaluable ecosystem analysis from Francois Chollet). These other things were cut for purposes of length and because we felt a mainstream reader (which is more Bloomberg's audience) would be able to understand the 'stars' figure most easily of all metrics.
Obviously I agree with you that this measure isn't definitive, but we felt it was relatively easy to understand. And the rest of the reporting we did for this story bore out the idea that TF has gathered an unusual level of both enthusiasm and commitment in a short amount of time.
One area where Google does seem to avoid this secrecy is in AI: both Google Brain and DeepMind are consistently publishing papers & in many cases putting them up on Arxiv pre-publication.
I'm an optimist in this area - people like Wait But Why and XKCD (eg - https://xkcd.com/thing-explainer/) - show that most complex subjects can be explained in simple language that most people with an average education can understand. (An aside: I recently wrote an article about the use of D-Wave quantum computers on Wall Street and that experience gave me a sense of how tremendously difficult it can be to simple summarize an inherently technical topic. I estimate it took about five hours of study to be able to write a couple of accurate sentences I felt comfortable with.) Whether the general public has the inclination to take the time to understand this stuff is another question entirely!
Hi there, author here -- mostly trying to be accessible. And I think in the tech community it is well understood that robotics is hugely difficult, but it seems like general public pretty much equates progress in software AI with progress in robots, which is clearly not the case. We thought it might be helpful to highlight this to people who aren't hugely technical.
Hiya. Reporter here. On the press briefing call Hassabis said that the single node version won 494 out of 495 games against an array of closed- and open-source Go programs. The distributed version was used in the match against the human and will be used in March. Distributed alphago used in the October match was about 170GPUs and 1200CPUs, he said.