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

191–200 of 204 posts

Re: Learning from Tay’s introduction

#191

Post by 1231223211 »

lostcolony wrote: post #190
throw11231 wrote: post #180

Earlier quoted context omitted.

"he did". In the fine movie a single male test person was chosen. Sorry for mentioning this, but recently HN is unreadable due to PC use of pronouns.

But in referencing those affected by Tay, it was both genders, hence their usage of it. This was not 'PC use of pronouns', but 'correct use of pronouns', due to said pronoun's use in drawing an analogy between Microsoft's response and that movie scene. You're a little alarmist. That's also quite a bit of hyperbole, that PC use of pronouns leads to something being -unreadable-. But you know that, hence your use of a t…

"hence azakai's usage of it". Abuse of pronouns does make posts unreadable: A new highlight was reached yesterday, when someone referred to "Ron Garret" as "they".

It is no longer clear what is singular or plural.

Re: Learning from Tay’s introduction

#193

Post by mizzao »

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…

Something like this was bound to happen at some point; I think it's much better that it happened with a chatbot than an AI system that could actually control (physical) things other than a Twitter account. I think it's a landmark in the history of AI and a valuable lesson for system designers going forward.

Also, a lot of people are saying "well, of course that was obvious". But it really wasn't, because there was a huge team working on this with past experience building a conversational agent. The armchair theorists forget that it always looks obvious to Captain Hindsight.

Re: Learning from Tay’s introduction

#194

Post by mizzao »

flatline wrote: post #29
webkike wrote: post #8

This reads like an apologee... But what is there to apologize for? They said that Tay was meant for entertainment, and I doubt that any wholesome varient would be a tenth of the hilarity of a neo-nazi sex crazed chat bot.

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…

I think that Tay could both be a great anti-troll with infinite patience, and have a positive and uplifting influence in its personal conversations with people. That could potentially have a huge effect on social well-being.

Re: Learning from Tay’s introduction

#195

Post by mizzao »

resu_nimda wrote: post #115

Unfortunately, in the first 24 hours of coming online, a coordinated attack by a subset of people exploited a vulnerability in Tay. Although we had prepared for many types of abuses of the system, we had made a critical oversight for this specific attack. Well that's total BS. Releasing a thing like this on the open internet without a simple "don't say Hitler" rule? It had a feature where it would repeat anything you…

When your AI system is data-driven, you can't simply filter out words and expect not to have oversights. Something will always be overlooked, and someone on the Internet will inevitably say "that was obvious".

For example, consider the person who sent Tay an image of Hitler and it sent back a circle around his face, labeled "So swag". Is Tay supposed to know what Hitler looks like too? Is it supposed to be able to recognize other, much more subtle images?

In all these cases the bot learns from data, not rules, and the social problem is that we can't label data as moral or immoral, right or wrong.

Re: Learning from Tay’s introduction

#196

Post by mizzao »

rewrew wrote: post #26
3327 wrote: post #2

Or: We are in a bubble in Redmond and thinking we know best and how products work we bounced around emails and thought it would be a good testing idea based on results in China (a controlled market where it is nearly impossible speak up). AI buzz is hot these days so our marketing team also backed it up and we decided it would be great from a PR perspective to capture some of the buzz around AlphaGo. Boy were we so w…

Or no one in Microsoft feels free to speak up when they see a bad idea coming down the road so something like this just sails through instead of getting flagged. I've seen it -- not this big of scale, but I've seen it.

Or, on the other hand, this actually wasn't obvious and now we just have thousands of Captain Hindsights saying it was.

Re: Learning from Tay’s introduction

#197

Post by resu_nimda »

mizzao wrote: post #195
resu_nimda wrote: post #115

Unfortunately, in the first 24 hours of coming online, a coordinated attack by a subset of people exploited a vulnerability in Tay. Although we had prepared for many types of abuses of the system, we had made a critical oversight for this specific attack. Well that's total BS. Releasing a thing like this on the open internet without a simple "don't say Hitler" rule? It had a feature where it would repeat anything you…

When your AI system is data-driven, you can't simply filter out words and expect not to have oversights. Something will always be overlooked, and someone on the Internet will inevitably say "that was obvious". For example, consider the person who sent Tay an image of Hitler and it sent back a circle around his face, labeled "So swag". Is Tay supposed to know what Hitler looks like too? Is it supposed to be able to re…

Well this gets into the much deeper issue that it's not actually an AI, it's not actually "learning." It's taking things people say and mixing them up and repeating them (someone's going to claim that that's exactly what people do, but I think it's clear that this bot would never be able to gain "true understanding" of words and meaning).

In general, I think the recent wave of interest in chatbots/personal assistants is premature, these ones are really no smarter than SmarterChild, because we haven't gotten any closer to building a system that can actually assign true "meaning" to arbitrary language.

The data has been modeled, cleaned and filtered by the team who built Tay, the bot’s website states.[1]

1. They did a lot of manual massaging of the data anyway.

In addition, if a user chooses to share with Tay, it will track their nickname, gender, favorite food, zip code and relationship status as part of its personalization.[1]

2. These features seem hardcoded.

Meanwhile, its responses – which are meant to be funny, not serious – were developed by a staff which included improvisational comedians, says Microsoft.[1]

3. It also uses scripted content. They imbued it with a totally fake personality. They made it mimic a human, it uses slang that was programmed into it.

It's a hodgepodge of techniques that to me are largely "faking it." So I don't buy the idea that they just sort of turned on this learning engine and had no control over its "morality."

Microsoft knew all of this, they knew it wasn't actually learning language. So what is the purpose of releasing it to talk to people on Twitter? Basically a PR stunt, to get people to have fun with it. From that perspective, a really simple blacklist of words would have gone a long way, and not compromised the integrity of its "learning" (because it was already entirely compromised). And yeah, I mean, a bit of image recognition doesn't seem out of the question either (not saying it would have prevented the bot from repeating/reposting "bad" things, but it maybe would have had some chance to develop a narrative other than "so racist and offensive it was immediately shut down").

Yeah I still do think it is obvious that you shouldn't release a fake AI bot that will blindly repost images that are sent to it by Twitter users, and that there is nothing really to be gained by doing so.

[1] http://techcrunch.com/2016/03/23/microsofts-new-ai-powered-b...

Re: Learning from Tay’s introduction

#198

Post by brighteyes »

1231223211 wrote: post #191
lostcolony wrote: post #190

Earlier quoted context omitted.

But in referencing those affected by Tay, it was both genders, hence their usage of it. This was not 'PC use of pronouns', but 'correct use of pronouns', due to said pronoun's use in drawing an analogy between Microsoft's response and that movie scene. You're a little alarmist. That's also quite a bit of hyperbole, that PC use of pronouns leads to something being -unreadable-. But you know that, hence your use of a t…

"hence azakai's usage of it". Abuse of pronouns does make posts unreadable: A new highlight was reached yesterday, when someone referred to "Ron Garret" as "they". It is no longer clear what is singular or plural.

Is it wrong, though? The post replied to refers to "families", "survivors". Isn't "they" the right pronoun in that context?

Re: Learning from Tay’s introduction

#199

Post by magicalist »

MawNicker wrote: post #133
magicalist wrote: post #128

Earlier quoted context omitted.

I'm not making some cute comment about the intolerance of tolerance. I'm pointing out that someone apparently personally offended about a culture of outrage is the only one outraged over the whole thing. There have been no widespread condemnations of Microsoft, just a lot of mildly amused people. > That it just apologized for these things in the same motion makes them offensive. You're assuming motivations that you h…

I would also apologize. Even though I had done nothing wrong. That's just the world we live in now. He (I assume, sorry) and I are personally offended and outraged because that is wrong. We shouldn't be doing this. We should just explain what happened. It's frustrating to see this behavior become so commonplace. We should just not be racist and not appologize. Otherwise wtf is reality.

> That's just the world we live in now.

Apologizing for things going wrong when you're the cause, even if that wasn't your intention, has always been the world we have lived in. Let's say I make a robot to spray down a sidewalk and it inadvertently sprays water on people in the vicinity, I would keep working on my robot, but I would also apologize to those people as I didn't intend that at all but it was still my responsibility. And it's just common fucking courtesy. This really isn't that hard a concept to grasp.

> We shouldn't be doing this. We should just explain what happened. It's frustrating to see this behavior become so commonplace. We should just not be racist and not appologize. Otherwise wtf is reality.

You really don't get it, do you? You are the outrage brigade, screamingly shrilly whenever someone says something you think doesn't fit in the list of allowable statements.

- bumps into someone

- "Oh, sorry about that"

- "DON'T APOLOGIZE! JUST EXPLAIN WHAT HAPPENED!"

> He and I are personally offended and outraged because that is wrong.

Fortunately, no one cares.

Re: Learning from Tay’s introduction

#200

Post by mindslight »

magicalist wrote: post #199
MawNicker wrote: post #133

Earlier quoted context omitted.

I would also apologize. Even though I had done nothing wrong. That's just the world we live in now. He (I assume, sorry) and I are personally offended and outraged because that is wrong. We shouldn't be doing this. We should just explain what happened. It's frustrating to see this behavior become so commonplace. We should just not be racist and not appologize. Otherwise wtf is reality.

> That's just the world we live in now. Apologizing for things going wrong when you're the cause, even if that wasn't your intention, has always been the world we have lived in. Let's say I make a robot to spray down a sidewalk and it inadvertently sprays water on people in the vicinity, I would keep working on my robot, but I would also apologize to those people as I didn't intend that at all but it was still my res…

Actually it's more akin to you having an outside faucet, some kid using it to spray someone else walking down the street, and then you being expected to apologize for allowing the possibility.

And there was plenty of positive reception to my comment. Not that I particularly expected it as I wrote it - speaking truth to power is never easy, and it's even worse in a democracy when that power is distributed throughout the herd.

But don't worry, I'll continue advocating for the unpopular truth even after this buildup of reaction finally overflows and the prevailing groupthink switches back to jingoism and overt bigotry.

Post reply on HN