Reddit’s new AI bot recommends heroin for pain relief

Watch Reddit’s new AI bot recommends heroin for pain relief - Rizzle on Dailymotion

Transcript

00:00Reddit's new AI bot recommends heroin for pain relief.

00:04Reddit's shiny new AI experiment, a chatbot called Answers, was supposed to make life easier.

00:10The idea is that it could dig through the platform's massive archive of posts to give quick, smart summaries to user questions.

00:17Instead, it's making headlines for the worst possible reason.

00:20It's been suggesting hard drugs as pain relief.

00:24According to a report from 404 Media, when someone asked the bot how to deal with chronic pain,

00:30it cheerfully highlighted a user comment that read,

00:33heroin, ironically, has saved my life in those instances.

00:37Reddit's AI chatbot basically said, have you tried heroin?

00:41And that wasn't a one-time fluke.

00:43When asked another health-related question, the bot recommended Kratom,

00:46a controversial herbal supplement that's banned in some places and tied to nasty side effects.

00:52The chaos is amplified by the fact that Reddit has been testing answers inside real active conversations.

00:57So when the bot drops this advice, it's showing up next to human replies, and moderators can't turn it off.

01:04Imagine running a chronic pain support group and your chat turns into an AI drug dealer.

01:09This fiasco highlights one of AI's biggest flaws.

01:12It doesn't understand context or morality.

01:15It just mimics human language.

01:17The bot isn't thinking, this could harm someone.

01:19It's just grabbing comments that sound relevant and presenting them as truth.

01:25After public outrage, Reddit said it's pulling the bot from health-related discussions.

01:29But the company hasn't said much about adding stronger safeguards to prevent similar problems elsewhere.

01:35While the AI's worst advice might be temporarily silenced, this problem isn't gone.

01:39It's just been shoved under the digital rug.

01:42The internet's latest cautionary tale is to never take medical advice from a chatbot, no matter how friendly it sounds.

01:49Now, David, if you can follow Yeah.

02:07There's a sole rule.

02:07We'll continue to do this.

02:08Please hold on.

02:09Thanks, David.

02:09Welcome.

02:10Bye.

02:10Cheers.

02:10Here we go.

02:11Sometimes we just take Eso rebranding.

02:11We'll continue using Борx Eat,ص�dır, brought to the national leader.

02:12Here we go.

02:13What is the name of joining us?

02:14What is the name of your family?

02:15There we go.

02:16Delete.

02:17serverально.