Hacker News Reader

Comment by Jeremy1026 | Hacker News Reader

Took longer than I expected for it to become transphoic/homophobic/racist. Given history of AI being left to its own devices.

View on HN

There's a lot more detail in a Reddit thread[1], but essentially an outage of the newer AI model forced them to downgrade to an older model, and the content moderation tools didn't catch the questionable jokes this time...

[1] https://www.reddit.com/r/LivestreamFail/comments/10v2r2e/com...

You should watch that clip. I'm curious what your thoughts are on it. I don't think it's being any of those things but it's making a joke about someone being those things and not getting laughs.

Yes! Remember Microsoft's chatbot Tay from 2016 that went super racist and forced MS to pull the plug?

https://money.cnn.com/2016/03/24/technology/tay-racist-micro...

Edit: A high-comment HN thread about the pulling:

https://news.ycombinator.com/item?id=11362187

Microsoft chatbot is taught to swear on Twitter 263 points; 228 comments https://news.ycombinator.com/item?id=11352307

Tay – Microsoft A.I. chatbot 154 points; 130 comments https://news.ycombinator.com/item?id=11346147

Can also go digging with this search range - https://hn.algolia.com/?dateEnd=1459468800&dateRange=custom&...

Right, I searched the Tay stories that way. I saw the second one but didn't link it because I was looking for stories from the aftermath, not the introduction. I missed the first one because the title didn't have Tay.

Most of the stories from your search don't fit either -- either referring to something else, or not having many HN comments (as I said I was looking for pre-edit).

The problem wasn't being unaware of the search feature.

Reading TFA, it looks like it only happened because they had trouble with ChatGPT and switched to Curie, a different AI text-generator that has less careful programming about switching into offensive concepts.

Reply on news.ycombinator.com