Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

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

futurism.com

Post reply on HN
Print view View original

111–120 of 201 posts

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

#111

Post by thewhitetulip »

fizzbatter wrote: post #87
chiph wrote: post #74

Earlier quoted context omitted.

Build a household AI, like JARVIS from Iron Man. Edit: Corrected.

Yup, this is my answer. I'm slowly building an in-house software suite .. because i can, i guess, and a big thing i'm wanting is to tie it all together with a human friendly administrative program. The main thing stopping me is NLP. Ideally, i want this offline only as i am unsure how much of my life i want to leave my home network.

I wanted to build an AI to plan my life. I buult a twitter clone to work on my phone only & I log my life in it. The AI would use my call logs & that App which I call Rants to do things like suggest calling someone I've not spoken to since a long time.

I haven't started yet, I'd be grateful if you would be open to discussing on this topic :-)

So far I didn't find any useful stuff on the internet

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

#112

Post by gohrt »

draugadrotten wrote: post #22

Back in 2007, mobile phones used a system called T9 from Nuance corp which was trained on a word corpus taken from IRC and similar chats. This caused all kinds of issues - the mobile phones would accept offensive words like "naziparking" but reject normal language like "world peace". Using reddit may lead to ... surprises. Source: http://spraktidningen.se/artiklar/2007/11/darfor-ar-din-mobi... Translated by Google: h…

T9 on phones goes all the way back to ~2000

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

#113

Post by gohrt »

JonnieCache wrote: post #77
LukeB_UK wrote: post #16

Earlier quoted context omitted.

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.

I remember FYAD doing the same thing to an eliza implementation that had learning back in the day, in like 2004. Plus ca change...

It's pretty well established that modern AI's main contribution is (a) massively larger datasets, (b) algorithms and technology to handle those massively larger datasets, (c) boring but important parameter tuning.

The core learning algorithms are not changing.

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

#114

Post by gohrt »

SnydenBitchy wrote: post #75
Cshelton wrote: post #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'.

“Witty” isn’t the first thing that comes to mind to describe reddit. I do look forward to hearing Siri telling me to get raped every time I summon it.

The top comments are witty. Like everything, 90% of Reddit is crap.

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

#115

Post by oldmanjay »

draugadrotten wrote: post #22

Back in 2007, mobile phones used a system called T9 from Nuance corp which was trained on a word corpus taken from IRC and similar chats. This caused all kinds of issues - the mobile phones would accept offensive words like "naziparking" but reject normal language like "world peace". Using reddit may lead to ... surprises. Source: http://spraktidningen.se/artiklar/2007/11/darfor-ar-din-mobi... Translated by Google: h…

I'm having a hard time imagining being sensitive enough that "naziparking" is offensive. Is any mention of the concept of nazis verboten?

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

#116

Post by gohrt »

bkanber wrote: post #3

This will be interesting. I'm sure they are, but I hope they'll be training the system on tone and sentiment alongside syntax. Reddit can get vitriolic and rude, insightful at times too, but once the system learns the syntax hopefully they'll be able to use sentiment analysis to weigh more strongly the polite conversion that occurs. Also interested to see how many memes this AI picks up. I also hope they are able to…

Filtering to only include long comments is simply way to drop a lot of cheaf and keep a lot of wheat, and have more internal context to help with analysis.

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

#117

Post by syllogism »

AlexCoventry wrote: post #106
syllogism wrote: post #6

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…

> I've applied for a grant from NVidia so I can fix that. A g2.xlarge is 65c/hour on AWS, FWIW. https://aws.amazon.com/ec2/pricing/on-demand/

That's honestly no way to get anything done...

Unless you're replicating someone else's thing exactly, You can't really get by with one training process. You want to be trying different things, and running a few samples of each configuration to account for random variation. I'm not even talking about decadent hyper-parameter sweeps to fine-tune. I'm talking like, how wide do I need my layers to be, what optimizers are good, how deep should I make the network, etc.

I want to be training 5-10 models at a time minimum. 20-30 would be much more productive. If I can only train one model at once, it's not really worth the effort — it's better to work on one of the other tickets for the library.

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

#118

Post by pc86 »

samstave wrote: post #108
AlexCoventry wrote: post #106

Earlier quoted context omitted.

> I've applied for a grant from NVidia so I can fix that. A g2.xlarge is 65c/hour on AWS, FWIW. https://aws.amazon.com/ec2/pricing/on-demand/

Yeah, for OD? thats 468/month... I'd use a spot instance and stop it whenever possible.

Presumably that's a decent chunk of what the grant is for?

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

#119

Post by minimaxir »

cotinoco wrote: post #103
minimaxir wrote: post #42

Earlier quoted context omitted.

SubredditSimulator was made by a Reddit Admin as a method of creating test data. It uses Markov chains for simplicity, which is not that exciting.

I wonder if it's this same code? https://github.com/reddit/reddit/blob/master/scripts/inject_...

No, the author has made it open source as a separate project: https://github.com/Deimos/SubredditSimulator

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

#120

Post by krashidov »

One of the things I like to do is play out a business to it's absurd maximum. What's the craziest possible future can I see for a company and its assets?

For Reddit, I like to imagine that it's basically the training data for all of the emotional and societal nuances that a human goes through.

Think about all of those stories that people post in ask Reddit that explain western norms and no nos. how to treat people with respect, when to call the police, how to communicate properly, etc.

Obviously we're far away from using the data to its full potential but one day I could see Reddit data to make our AIs more relatable and human like.

Post reply on HN