Live data from Hacker News

OpenAI is Using Reddit to Teach An Artificial Intelligence How to Speak

futurism.com

11–20 of 201 posts

Re: OpenAI is Using Reddit to Teach An Artificial Intelligence How to Speak

#11

The DGX-1 is available for a cool $129k: http://www.nvidia.com/object/deep-learning-system.html Correct me if I'm wrong, but I think it's basically a couple hundred NVIDIA 10-series cards strapped together with a full custom NVIDIA software stack.

A DGX-1 box has 8 Pascal GPUs. The reason it costs a lot more than 8 GTX 1080s is the remarkable interconnect and memory bandwidth.

Re: OpenAI is Using Reddit to Teach An Artificial Intelligence How to Speak

#12
As a frequent Redditor, this AI is going to be very witty.

They should limit it to top comments only, and for training, you might as well assume 90% of top comments are sarcastic/tongue in cheek. Or let a user dial the sarcasm/wittiness/seriousness as they want it, kind of like TARS from 'Interstellar'.

Re: OpenAI is Using Reddit to Teach An Artificial Intelligence How to Speak

#13

The Reddit comment corpus is an awesome dataset. There's relatively little mark-up to scrub out, low duplication, good metadata, and a variety of topics. We used it to train a syntax-enriched word2vec model. Write up and demo: https://explosion.ai/blog/sense2vec-with-spacy Btw, the above was run on CPU in a couple of days, because spaCy doesn't use GPUs yet. I've applied for a grant from NVidia so I can fix that. If…

> But when we ran the model on more data, and it was gone and soon forgotten. Just like Carrot Top.

"it was gone" meaning the association between Carrot Top and Kate Mara? So after better training, who is now most_similar(['Carrot_Top|PERSON'])?

EDIT: RTFA, used the interactive demo. most_similar() is now a category I would describe as "actors/comedians popular in the 90's: "Bill Murray, Gary Busey, David Spade, Charlie Sheen, Ashton Kutcher, Chris Farley".

Re: OpenAI is Using Reddit to Teach An Artificial Intelligence How to Speak

#14

As a frequent Redditor, this AI is going to be very witty. They should limit it to top comments only, and for training, you might as well assume 90% of top comments are sarcastic/tongue in cheek. Or let a user dial the sarcasm/wittiness/seriousness as they want it, kind of like TARS from 'Interstellar'.

That was first thought. Curious How does NN or any other model capture sentiment ?

Re: OpenAI is Using Reddit to Teach An Artificial Intelligence How to Speak

#15
post #14

As a frequent Redditor, this AI is going to be very witty. They should limit it to top comments only, and for training, you might as well assume 90% of top comments are sarcastic/tongue in cheek. Or let a user dial the sarcasm/wittiness/seriousness as they want it, kind of like TARS from 'Interstellar'.

That was first thought. Curious How does NN or any other model capture sentiment ?

It's a tough problem. You need a lot of context data. Reddit is definitely a good set to train on.

Re: OpenAI is Using Reddit to Teach An Artificial Intelligence How to Speak

#16

Didn't msft do the same thing with twitter and end up with racist bot? I am not sure how this will turn out.

Microsoft's bot (Tay) learned as people talked with it. People took advantage of that and basically attacked it with racist things which meant it ended up learning to be a racist.

Re: OpenAI is Using Reddit to Teach An Artificial Intelligence How to Speak

#17

The Reddit comment corpus is an awesome dataset. There's relatively little mark-up to scrub out, low duplication, good metadata, and a variety of topics. We used it to train a syntax-enriched word2vec model. Write up and demo: https://explosion.ai/blog/sense2vec-with-spacy Btw, the above was run on CPU in a couple of days, because spaCy doesn't use GPUs yet. I've applied for a grant from NVidia so I can fix that. If…

> But when we ran the model on more data, and it was gone and soon forgotten. Just like Carrot Top. "it was gone" meaning the association between Carrot Top and Kate Mara? So after better training, who is now most_similar(['Carrot_Top|PERSON'])? EDIT: RTFA, used the interactive demo. most_similar() is now a category I would describe as "actors/comedians popular in the 90's: "Bill Murray, Gary Busey, David Spade, Char…

https://demos.explosion.ai/sense2vec/?word=carrot%20top&sens...

Weirdly there's a bug that's dropped PERSON from the sense list. Fixing.

Edit: Fixed.

Edit2: Ah this is super misleading atm. I'll have a think about how to do this better. auto is case insensitive, but if you specify a sense, it's case sensitive. So you need to do "Carrot Top" and set PERSON. Btw, contrast with carrot top "NOUN".

Re: OpenAI is Using Reddit to Teach An Artificial Intelligence How to Speak

#20
How does the team plan to address the issues faced by Microsoft's twitter chatbot Tay [0], which had racist inputs and in turn gave similar responses? While I don't know how recent the corpus is, the majority of reddit speaks like and holds the views of college-aged white males, and many of the things said on reddit have been deplorable. It'd be a shame if OpenAI pooled all that computing power into training on a bad data set, resulting in an AI that regurgitates memes and random references in response to anything.

[0]: http://www.theverge.com/2016/3/24/11297050/tay-microsoft-cha...

Post reply on HN