I haven't had a chance to listen to the audio yet, but I did read your piece on this late last week. I'm not a programmer but I write a lot about AI for my employer, a business consulting company, and with a colleague I write our company's rules for using AI to help generate content. I think the problem is that in their haste to roll out generative AI, OpenAI, Google, and others focused more on its ability to follow grammar rules to create human-sounding output without any kind of bullshit detector. AI cannot distinguish accurate content from bullshit content at this point. All it has been built to do now is scrape everything, including all the bullshit people have posted about you in various chatrooms, etc. AI tries to compile it all in a way that makes grammatical sense without bearing on reality. Again, I'm not an AI expert or enthusiast, as someone who's spent 30 years writing the old-fashioned way and have no plans to turn my talents over to AI. But it seems to me Open AI's, Google's, and others' time would have been better spent up front creating an AI that could evaluate the validity of web addresses and IP addresses as a first screen, at least. Random comments idiots make should not be included in what it hoovers up into its algorithms, because in the process of turning into human language it prioritizes making grammatical sense, not generating truth.