Google’s generative AI gets “can cats be vegan” right at first….
but THEN decides to throw in a completely wrong opposing opinion, which some people who REALLY want a vegan cat will absolutely cling to.
I do think that this is somewhat deeper than a genAI problem, and a situation where there really is a misinformation campaign on the vegan side and it's not adequately matched by good information. Google's old algorithms would also fail.
every time i've replicated a stupid google AI response, the SEO problem is right under the surface. "is raw milk good for immunity" gave me a bad answer... summarized from the first search result, a pro-raw-milk site, which i also got with AI turned off
Yeah - like, that's what's kind of strange to me: Google's big AI push does not seem like an attempt to synthesize an enormous pool of training data, it seems like it's just "Generate a summary of the top three results"
yeah, i've found myself just scrolling down to the first link while the AI loads. the "show an excerpt from the first link" response is usually faster and more concise writing
a big part of the problem now though is that the summaries have the implicit imprimatur of Google because they are not only supposed to be fancy, AI-generated answers but they are right there at the top of the results page in the same font as the rest of the official Google stuff.
even people casually attuned to the wisdom that internet search results are unreliable might not immediately grok that Google would make obviously bad information so prominent and official-looking