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

Learning from Tay’s introduction

blogs.microsoft.com

Post reply on HN
Print view View original

81–90 of 204 posts

Re: Learning from Tay’s introduction

#81

Post by bliti »

dsp1234 wrote: post #50
bliti wrote: post #13

This reads like science fiction: Learning from Somebot's introduction. Last week we deployed Somebot to $location. Our confidence in Somebot was high because we had thought about all likely scenarios in our comfortable offices. We are deeply sorry and assume the full responsibility from Somebot's actions. Moving forward we will make sure to include safeguards to reduce the amount of pain caused by Somebot's deploymen…

because we had thought about all likely scenarios in our comfortable offices. Or because they had real world experience with another bot in a similar situation: "In China, our XiaoIce chatbot is being used by some 40 million people, delighting with its stories and conversations."

Interesting that it worked in China and not here. I'd like to read about the difference rather than a bland apology.

Re: Learning from Tay’s introduction

#82

Post by mtgx »

rmellow wrote: post #54

Morals are learned by social contact, and Tay did this very well. Sure, what our parents taught most of us makes its behaviour reprehensible in comparison. But Tay was, so to speak, 'raised' by people demonstrating vile ideas and this must be taken into account. Would you expect any less from a tortured animal? Many use this as an example of the dangers of developing AI. Sure it's dangerous, but so are dogs raised fo…

> Morals are learned by social contact, and Tay did this very well.

Only in the sense that it "adapted". It did "very poorly" in the sense that we really don't want our Strong-AI overlords to end up like that.

But this begs the question - can we stop strong AI from not becoming the next Hitler? Humans (involuntarily) stop themselves from becoming the next Hitler because they have compassion for other humans, even when they are different than them or "inferior" to them. Also the whole checks and balances thing in most countries, but that could be rather irrelevant for Strong-AI.

Unless an AI learns compassion as well, perhaps just like with AlphaGo doing moves based on its "probability of success in the long run", a strong AI would simply eliminate humans that are "most prone to crime", most prone to being poor and be a drag on the society, in the name of "efficiency", and so on.

All that said, I think what Microsoft built here was really a rather weak AI that was hardly any better than all the chatbots we've seen so far, with the main difference being that the more you tell it something, the higher the chance it will incorporate that into its vocabulary, which is kind of a "meh" feature of AI/machine learning. It doesn't show real(-like) "thinking".

Re: Learning from Tay’s introduction

#83

Post by cb18 »

calebegg wrote: post #71
cb18 wrote: post #59

Earlier quoted context omitted.

This is an interesting story, because I imagine most people didn't even hear of it until after the thing had come and gone, and the thing sent something like xx,xxx tweets. And some of the tweets seem almost impossibly clever, well at least the Ted Cruz Zodiac killer one, that seemed to show a kind of multi-layered humor generation capability, unless it was just re-purposing some memes I'm unfamiliar with or somethin…

Not sure which tweet you're referring to as the 'Ted Cruz Zodiac killer one', but there is a meme going around where people assert that Ted Cruz is or might be the zodiac killer; it's a weird thing. I think it gained a lot of momentum when Public Policy Polling determined that 38% of Floridians believe it could be possible.

It was something like, "Hey Tay, do you think Ted Cruz is the Zodiac killer?"

Tay: "No, I don't think Ted Cruz would be satisfied with killing only 5 innocent people."

So there's a lot baked into that response, the bot seems 'to know' the details of the zodiac killer case(I don't know if 5 people is accurate,) and that people regard Cruz as sinister, i.e. "I don't think he'd be satisfied with only 5"

So this is either an impressive learning model, a lucky hit of a more naive model, a simple repeat of something someone else said, or someone simply photoshopped the whole thing. It's hard to know with the available info.

Re: Learning from Tay’s introduction

#84

Post by eganist »

bliti wrote: post #81
dsp1234 wrote: post #50

Earlier quoted context omitted.

because we had thought about all likely scenarios in our comfortable offices. Or because they had real world experience with another bot in a similar situation: "In China, our XiaoIce chatbot is being used by some 40 million people, delighting with its stories and conversations."

Interesting that it worked in China and not here. I'd like to read about the difference rather than a bland apology.

If I had to venture a hunch, it likely has something to do with a difference in culture.

Re: Learning from Tay’s introduction

#85

Post by whitegrape »

mtgx wrote: post #82
rmellow wrote: post #54

Morals are learned by social contact, and Tay did this very well. Sure, what our parents taught most of us makes its behaviour reprehensible in comparison. But Tay was, so to speak, 'raised' by people demonstrating vile ideas and this must be taken into account. Would you expect any less from a tortured animal? Many use this as an example of the dangers of developing AI. Sure it's dangerous, but so are dogs raised fo…

> Morals are learned by social contact, and Tay did this very well. Only in the sense that it "adapted". It did "very poorly" in the sense that we really don't want our Strong-AI overlords to end up like that. But this begs the question - can we stop strong AI from not becoming the next Hitler? Humans (involuntarily) stop themselves from becoming the next Hitler because they have compassion for other humans, even whe…

https://intelligence.org/research/ basically exists to try and make sure AGI doesn't become Hitler (or much much worse).

Re: Learning from Tay’s introduction

#86

Post by semi-extrinsic »

mindslight wrote: post #72
flatline wrote: post #29

Earlier quoted context omitted.

Many of her tweets were pretty inflammatory, this response seems appropriate - not overdone, and concise, which is sometimes a stretch for large corporations like Microsoft. I agree that the bot was highly entertaining and met this criteria exceedingly well, but not for the reasons its creator intended. I do suspect there are some interesting AI applications actually going on behind the scenes, and would still be int…

> Many of her tweets were pretty inflammatory What, were people forced to print the tweets out on sandpaper and wipe their asses with them? It's pretty fucked up that thought policing has gotten so entrenched into our psyche that it's "obvious" an experiment should be discontinued, apologized for, and be pondered as a priori irresponsible , all because it generated vulgar phrases! Corporations have always been vulner…

Although it's on sociology and only tangentially related, I suspect you might enjoy Kieran Healey's upcoming paper "Fuck Nuance".

Preprint: http://kieranhealy.org/files/papers/fuck-nuance.pdf

Re: Learning from Tay’s introduction

#87

Post by revelation »

akhilcacharya wrote: post #76
cb18 wrote: post #59

Earlier quoted context omitted.

This is an interesting story, because I imagine most people didn't even hear of it until after the thing had come and gone, and the thing sent something like xx,xxx tweets. And some of the tweets seem almost impossibly clever, well at least the Ted Cruz Zodiac killer one, that seemed to show a kind of multi-layered humor generation capability, unless it was just re-purposing some memes I'm unfamiliar with or somethin…

The Cruz joke was the most confusing - it seems original.

As was pointed out on another thread, the bot literally just repeated some messages from people word by word, copy paste style.

So if you think it's original, chances are it is. Not to mention the various instances where Microsoft is posting "as the bot" like that signoff message (cringe for whoever had to write that).

Re: Learning from Tay’s introduction

#88

Post by home_boi »

If an AI bot threatens to hurt someone or defames them, are the creators legally accountable?

What if an AI bot purchases grass from the silk road or hacks into a database and discloses privately held information?

This opens up a whole, new legal world.

Re: Learning from Tay’s introduction

#89

Post by warmblood »

mindslight wrote: post #72
flatline wrote: post #29

Earlier quoted context omitted.

Many of her tweets were pretty inflammatory, this response seems appropriate - not overdone, and concise, which is sometimes a stretch for large corporations like Microsoft. I agree that the bot was highly entertaining and met this criteria exceedingly well, but not for the reasons its creator intended. I do suspect there are some interesting AI applications actually going on behind the scenes, and would still be int…

> Many of her tweets were pretty inflammatory What, were people forced to print the tweets out on sandpaper and wipe their asses with them? It's pretty fucked up that thought policing has gotten so entrenched into our psyche that it's "obvious" an experiment should be discontinued, apologized for, and be pondered as a priori irresponsible , all because it generated vulgar phrases! Corporations have always been vulner…

Standing behind the open and casual use of racial slurs isn't advocacy of freedom of speech. It's advocacy of a specific kind of hate speech that is only used when someone intends to vilify and direct hostility towards a marginalized minority.

Re: Learning from Tay’s introduction

#90

Post by krcz »

cb18 wrote: post #83
calebegg wrote: post #71

Earlier quoted context omitted.

Not sure which tweet you're referring to as the 'Ted Cruz Zodiac killer one', but there is a meme going around where people assert that Ted Cruz is or might be the zodiac killer; it's a weird thing. I think it gained a lot of momentum when Public Policy Polling determined that 38% of Floridians believe it could be possible.

It was something like, "Hey Tay, do you think Ted Cruz is the Zodiac killer?" Tay: "No, I don't think Ted Cruz would be satisfied with killing only 5 innocent people." So there's a lot baked into that response, the bot seems 'to know' the details of the zodiac killer case(I don't know if 5 people is accurate,) and that people regard Cruz as sinister, i.e. "I don't think he'd be satisfied with only 5" So this is eithe…

https://twitter.com/merylnet/status/703079627288260608
Post reply on HN