recentpopularlog in


« earlier   
RT : Can we teach robots ethics? As we become more reliant on , an interesting point about human characteristics…
AI  from twitter
11 hours ago by adrianh
AlphaGo Zero: Learning from scratch | DeepMind
From Benedict Evans's newsletter: "Google's Deep Mind has reworked AlphaGo dramatically. Last year it won Go after its learning process was seeded with expert human games to observe, but the new version doesn't need any such data - all they gave it was the rules, and it was left to train itself by playing millions of games against itself. This new version beats the old one 100:0. As before, it generates moves and strategies that look nothing like any human expert has ever thought of. Practical implication: what fields are narrow enough that we can tell an ML system all the rules and let it know what score it's achieving, but deep enough that it could do much better than a person has ever done? Hence, where can this model achieve new knowledge?"
deepmind  alphago  ai  machinelearning  bigdata  benediktevans 
11 hours ago by WBedutech
Vertex.AI - Announcing PlaidML: Open Source Deep Learning for Every Platform
this could be a big deal, accelerated ML on any opencl device (including my humble macbook air), mac port coming soon
ai  machinelearning  python 
13 hours ago by aparrish

Copy this bookmark:

to read