AI Hallucinating Code Outputs Garbage....
quietus writes:
About a month ago, a developer notified The Register (El Reg, for the conoscenti) that part of the documentation of an AI powered infrastructure-as-code tool was turning up on the Internet. Problem: the documentation itself was written by an LLM and, as one has come to expect, it sometimes hallucinated. In layman terms: it shoveled bullshit.
The developer -- Pete Nykanen -- had opened an issue with the company on Github, but no action was taken.
Lo and behold, El Reg went on a fishing expedition and found that the story was true. What's more, the bogus content ended up number one in Google's search results, and pretty high-up almost everywhere else too. Intrigued, El Reg investigators asked both the Google and the original AI polluter for comment, which led to some interesting quotes.
The original polluter noted that it has "taken steps to remove more than half (almost two thirds) of AI Answers, and we plan to continue to ensure that these AI answers are complementary to our existing documentation." They even planned to refer to existing APIs and documentation in the answers, and testing generated code was on the to-do list too.
How they're going to do that -- given their proven ability in planning ahead and stuff -- is probably merely a question of filling in the blanks when asking the question to their own AI engine, one supposes.
Google was a bit slower in responding -- by about a month or so -- but helpfully noted that "Low value content that's created at scale to manipulate Search rankings is spam, however it is produced", and that recent updates to its tech "reduced low quality, unoriginal content on Search by 45 percent, and aim to tackle unhelpful content that's designed to rank well in Search."
Unfortunately, it looks like them updates missed the mark: at the time of writing this sub, the original content is still number one at Google.
Read more of this story at SoylentNews.