gdb

11 ms·

91.
OpenAI Five(blog.openai.com)
646 points by gdb 8y ago|124 comments
92.
Improving Language Understanding with Unsupervised Learning(blog.openai.com)
191 points by gdb 8y ago|18 comments
93.
by gdb 8y ago
Updated the post :).
94.
AI and Compute(blog.openai.com)
347 points by gdb 8y ago|130 comments
95.
Evolved Policy Gradients(blog.openai.com)
48 points by gdb 8y ago|4 comments
96.
by gdb 9y ago
Yep!
97.
OpenAI Retro Contest(contest.openai.com)
197 points by gdb 9y ago|48 comments
98.
OpenAI Retro Contest(blog.openai.com)
1 points by gdb 9y ago|0 comments
99.
New robotics environments for OpenAI Gym(blog.openai.com)
1 points by gdb 9y ago|0 comments
100.
Discovering types for entity disambiguation(blog.openai.com)
85 points by gdb 9y ago|6 comments
101.
OpenAI is hiring infrastructure engineers(jobs.lever.co)
1 points by gdb 9y ago
102.
Learning a hierarchy(blog.openai.com)
267 points by gdb 9y ago|39 comments
103.
Generalizing from simulation(blog.openai.com)
3 points by gdb 9y ago|0 comments
104.
by gdb 9y ago
Aw, that was my one contribution to the video :)! What kind of music would you have preferred instead?
105.
Competitive Self-Play(blog.openai.com)
374 points by gdb 9y ago|138 comments
106.
by gdb 9y ago
(I work at OpenAI.) Correct, we've been playing a number of pros using the same bot played at TI. We do have a stronger version which is just two days more of training (gets a 70% win rate vs the one at TI), but haven't seen a nee
107.
by gdb 9y ago
(I work at OpenAI.) TensorFlow by default is built with denormals off (ftz=true), so denormals aren't relevant for the applications we're interested in. We have updated the post to indicate this — thanks for the feedback!
108.
Building the Dota 2 system(blog.openai.com)
27 points by gdb 9y ago|1 comments
109.
by gdb 9y ago
See the opening video ( https://www.youtube.com/watch?v=wiOopO9jTZw&t=268 ), where I mention that we "coach" it on what we think is good and bad. All the heavy lifting is self-play (which I would characterize as
110.
by gdb 9y ago
(I work at OpenAI.) We'll have another blog post coming in the next few days. But as a sneak peek: we use self-play to learn everything that depends on an interaction with the opponent. Didn't need to with those that don't (e
111.
OpenAI is hiring Large-Scale Reinforcement Learning engineers(openai.com)
1 points by gdb 9y ago
112.
by gdb 9y ago
(I work at OpenAI.) Great project. We've found that the VNC Universe environments are hard for today's RL algorithms primarily due to the their async nature. We're currently working on a new set of Universe environments witho
113.
by gdb 10y ago
Yep weights will be up soon! EDIT: in fact, weights were up at launch: https://github.com/openai/generating-reviews-discovering-sen...
114.
by gdb 10y ago
We did not use the star ratings.
115.
by gdb 10y ago
Thanks for the feedback — added context to that sentence to make it more clear! The main interesting thing is that none of the Amazon data was labeled, while the 232 labeled examples were.
116.
Unsupervised sentiment neuron(blog.openai.com)
603 points by gdb 10y ago|126 comments
117.
by gdb 10y ago
While the application is indeed a joke, the technique for sim-to-real transfer (and the robotic system in the post) is very real: https://arxiv.org/abs/1703.06907 :).
118.
Evolution Strategies(blog.openai.com)
8 points by gdb 10y ago|0 comments
119.
by gdb 10y ago
(I am Greg.) Starting something new is incredibly hard. The default is that your company never even forms, and it's on you to overcome the activation energy. I don't think I'd have the mental fortitude to stick that out alone
120.
by gdb 10y ago
> Greg is the guy you need. Not necessarily the guy you want to be. (I am Greg, and am reasonably happy to be me.) I think this reply is true. But even that aside, the startup philosophy that guides me is variance maximization. So yes, w

More ›