undefined | Better HN

0 pointsshkkmo4y ago0 comments

Not only did Alexa retrieve the wrong content (instead of a list of challenges to do, Alexa sourced from a list of dangerous challenges you shouldn't do), Alexa then stripped off the entire context of the warning about not doing the challenge because it was dangerous, and instead added a timer.

So the issue here is less about needing to sanitize the internet for kids, and more about how AI struggles to categorize and summarize content and the dangerous effects that can have.

0 comments

encryptluks24y ago

You overestimate what AI means in terms of marketing. If a site owner doesn't markup their site properly, then that could also be the cause of issues like this. Do you really expect Alexa to be "smart" enough to understand human logic like an actual human, or do you not see that AI can basically mean Alexa goes to the web, does a search and parses markup from a website to the best of its ability and they call it AI? Some people care about the technicalities and others just want it to work perfectly, but I think the people that care about the technicalities are more realistic than just wanting something to be the way they think it should be.

shkkmoOP4y ago

> You overestimate what AI means in terms of marketing.

Not really, you are making huge assumptions about me here. I don't expect Alexa to be smart. I do expect that if a company is going to scrape other people's content to use in their service, it behooves them to do that scraping properly and to not use automation to do it if their automation isn't up to snuff.

hakfoo4y ago

The right answer is that Alexa (especially in a "Kids" mode) should only pull from a white-listed, manually vetted data pool. It wouldn't find the list of dangerous challenges in the first place.

In the worst case, this means, oh no, they might need to find human moderators to approve content for the pool, or even custom create content. They can afford it.

j / k navigate · click thread line to collapse