wuthefwasthat

4 ms·

1.
by wuthefwasthat 8y ago
Hi! OpenAI employee and paper co-author here, speaking entirely on my own (nobody else in the company knows I'm posting this). Occam's razor is a great principle, and I completely see how it might point towards your viewpoint.
2.
by wuthefwasthat 8y ago
Yeah, there are definitely still places the samples fall short! Keep in mind we're still using very naive sampling techniques. RE Winograd: WNLI is different, see https://arxiv.org/pdf/1804.07461.pdf
3.
by wuthefwasthat 8y ago
Those samples are from the large model (GPT-2)! Regarding memorization vs. generalization, see our paper for more analysis.
4.
by wuthefwasthat 8y ago
Thank you! We've released 500 random unconditional samples from GPT-2 at https://github.com/openai/gpt-2/blob/master/gpt2-samples.txt
5.
by wuthefwasthat 8y ago
Never thought I'd see my github repo cited in a DeepMind/Brain paper. I still have SOTA for >2 player by a large margin :) I'd be pretty impressed if they beat my 5p strategy without Hanabi-specific stuff, but I assume th
6.
by wuthefwasthat 9y ago
Yeah, I'd add it if it helps other developers get started with the code base. Happy to chat with you about it, though, too! Feel free to email me. That said, I'm not sure there's anything particularly interesting or good ab
7.
by wuthefwasthat 9y ago
Thanks! The lag was for initial page load, and is just a historical footnote, probably. Here's what I dug up from past chats: In May 2015, for a 250k bullet document, it used to take more than 5 minutes. Apparently by July, it was do
8.
by wuthefwasthat 9y ago
It basically comes for free if you implement the tree operations efficiently :)
9.
by wuthefwasthat 9y ago
Let me know how it goes!
10.
by wuthefwasthat 9y ago
I made ? a bit more discoverable. Thanks for the feedback! It's pretty easy to experiment with a new data backend, as long as it's easy to implement GET and SET for a key-value store on top of it. It took me less than an hour to
11.
by wuthefwasthat 9y ago
You're welcome! Let me know what you end up thinking. I'm quite curious, since I came from a similar situation as you, but went less far with org mode (I've tried the spacemacs org-mode layer but didn't like it that mu
12.
by wuthefwasthat 9y ago
It is possible to customize the shortcuts, though it's a bit cumbersome. In Settings under the Hotkeys tab, you can download a JSON file describing your mappings, edit it, then re-upload it.
13.
by wuthefwasthat 9y ago
Those both work for me. What browser are you using? (Are you using vimium or vimperator maybe?)
14.
by wuthefwasthat 9y ago
Author here! Didn't see this coming - time to update all the documentation :) Thanks for the comments, everyone, and feel free to ask questions here. Some background: I started working on Vimflowy after a conversation with a friend.
15.
by wuthefwasthat 10y ago
You say it's impossible to prove a universal negative? I'd love to see a proof of that
16.
A homomorphic encryption scheme explained to high schoolers(wuthefwasthat.github.io)
1 points by wuthefwasthat 10y ago|0 comments
17.
by wuthefwasthat 10y ago
aliases for opening files, which i find quite useful # open file by name op = "!f() { $(git editor) $(git ls-files | grep -i \"$@\"); }; f" # open currently edited (unstaged) files eop = "! $(git editor)
18.
by wuthefwasthat 12y ago
we just added some official base instances. check them out! http://www.terminal.com/explore#query=username%3Aterminal