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

Facebook uses 1.5B Reddit posts to create chatbot

bbc.com

Post reply on HN
Print view View original

61–70 of 235 posts

Re: Facebook uses 1.5B Reddit posts to create chatbot

#61

Post by threatofrain »

r00fus wrote: post #57
threatofrain wrote: post #50

Earlier quoted context omitted.

Reddit a dumpster fire? Some of Reddits technical communities have a way higher signal than HN, like Reddit’s math or rust community. Reddit math is more curious about math than HN is about tech.

signal and empathy seem orthogonal traits to me.

Reddit's best big technical communities are better on this regard too. Their moderation is more transparent and dependent on volunteerism, and somehow even the more serious communities squeeze in more humor.

Also, HN gives you the vibe such that you'd wish to argue about the orthogonality of empathy and signal? As opposed to HN feels like SO?

Re: Facebook uses 1.5B Reddit posts to create chatbot

#63

Post by wallacoloo »

r00fus wrote: post #57
threatofrain wrote: post #50

Earlier quoted context omitted.

Reddit a dumpster fire? Some of Reddits technical communities have a way higher signal than HN, like Reddit’s math or rust community. Reddit math is more curious about math than HN is about tech.

signal and empathy seem orthogonal traits to me.

Do they actually though? If you have empathy for the people in your community, are you not more likely to contribute more meaningfully to that community (I.e. more signal)?

Re: Facebook uses 1.5B Reddit posts to create chatbot

#65

Post by unixhero »

mrweasel wrote: post #51
gundmc wrote: post #44

From the cherry-picked conversation section of the research paper [1]. > Human: Hi! > Bot: Hi! How are you? I'm a vegan, so I don't eat any animal products. Even AI vegans reinforcing stereotypes of being overly-forthcoming. [1] - https://arxiv.org/pdf/2004.13637.pdf

Well, training it on Reddit data it's going to take on a vegan, pan-sexual, Bernie Sanders supporting, urbanite, Tesla driving, artist / cam-girl persona.

Cam-girl hahahaha

Re: Facebook uses 1.5B Reddit posts to create chatbot

#66

Post by unixhero »

sheeshkebab wrote: post #60

I’ve been developing some chat bots recently professionally and, honestly, I don’t get the point of these kinds of projects, especially done by companies like Facebook - sure you could create some text generator based on tokens, but it serves no purpose. Give me a trained bot that can extract specific things in various different ways users express them (without me creating dumb questionnaires), match across thousands…

It's one step further. Scientific research is what it is.

Re: Facebook uses 1.5B Reddit posts to create chatbot

#67

Post by nitwit005 »

ben_w wrote: post #55
gundmc wrote: post #44

From the cherry-picked conversation section of the research paper [1]. > Human: Hi! > Bot: Hi! How are you? I'm a vegan, so I don't eat any animal products. Even AI vegans reinforcing stereotypes of being overly-forthcoming. [1] - https://arxiv.org/pdf/2004.13637.pdf

The only vegans you ever know to be vegans are those who bother to tell you, so it shouldn’t be surprising.

Work lunches, weddings, and so forth are a control for that. My experience is you always know who the vegans are in advance.

Although I have learned from that that you need to hover around the vegan dishes to stop the non-vegans from eating it all before the vegans get to it, or order far more than should be needed.

Re: Facebook uses 1.5B Reddit posts to create chatbot

#69

Post by stephenroller »

cbuskilla wrote: post #45

if anyone wants to try it, I'm hosting the small model here: https://cocohub.ai/blueprint/blender_pv1/about

Hi there. One of the original authors of the BlenderBot paper here. This cocohub website is not what we released. It appears that cocohub has removed our safety layer and provided other modifications to the bot. We released our model open source, along with specific instructions on interacting with it safely: https://parl.ai/projects/recipes/

It's important to note that dialogue safety is a very important and nuanced topic, and we did our best to release a safety system attached to the model. Our system is not perfect though, and that is why BlenderBot was released as a research project for furthering the state of Artificial Intelligence, and is not meant for production purposes.

I would also mention that the blender small model significantly underperforms compared to the larger models released with the paper, and encourage everyone to try our best models, not our small one.

Re: Facebook uses 1.5B Reddit posts to create chatbot

#70

Post by cbuskilla »

xaedes wrote: post #53
cbuskilla wrote: post #45

if anyone wants to try it, I'm hosting the small model here: https://cocohub.ai/blueprint/blender_pv1/about

Thank you! But to be honest Blender is kinda underwhelming. I had better conversations with alice derivates. Blender feels bland, like a dozen different responses with only some words from my text inserted.

Sure! It is the 90M params models and they trained models up to almost 10B params so I guess it gets better with the size (Didn't try way too expensive).

And I agree about the alice derivates mitzuku is nice without doing anything fancy.

Post reply on HN