News for hackers. Stuff that matters.
Re:Skynet Today – accessible and informed coverage of AI hyp... by andreyk on Monday April 09, 2018 @03:46PM (#16796316)
Creator here! I and my collaborators are very excited to be doing this project, as AI hype has really gotten out of hand (even right here on HN sometimes). Yann LeCun endorsing us a few days back meant a great deal. Feel free to ask anything!
Re:Skynet Today – accessible and informed coverage of AI hyp... by robotsandml on Monday April 09, 2018 @11:52PM (#16798873)
What do you think are the main drivers for the hype? (Some that come to mind -- lack of understanding of AI, fear-mongering, deluge of published papers in academia without peer review...) And how are ways we can more realistically understand the state of AI/potential of it?
Re:Skynet Today – accessible and informed coverage of AI hyp... by p1esk on Monday April 09, 2018 @04:26PM (#16796674)
Which direction in AI is most promising towards AGI, in your opinion (regardless of the hype it receives)?
Re:Skynet Today – accessible and informed coverage of AI hyp... by andreyk on Monday April 09, 2018 @05:00PM (#16796956)
I'd say a hybrid of reinforcement learning, meta-learning, neural net structure evolution, and Google's sparsely gated outrageously large neural nets (https://arxiv.org/abs/1701.06538).

I have a new long essay on RL and meta-learning that addresses some of this coming out soon, keep an eye out!