Google AI Overview made up an elaborate story about me

Can we stop conflating LLM models with the companies that created them? It's "…Gemini made up…". Do we not value accuracy? It'd be a whole different story if a human defamed you, rather than a token predictor.

Why shouldn’t they be conflated? Google made the LLM, it is responsible for the LLMs output.

I mean, no, I don’t think some Google employee tuned the LLM to produce output like this, but it doesn’t matter. They are still responsible.

LLMs have no sovereignty, identity, or accountability - they are computer programs.

We do not blame computer programs when they have bugs or make mistakes - we blame the human being who made them.

This has always been the case since we have created anything, dating back even tens of thousands of years. You absolutely cannot just unilaterally decide to change that now based on a whim.

The HN headline I commented on was "Turns out Google made up an elaborate story about me". Someone (admin?) then added "AI Overview". My comment doesn't apply to the current headline. I got what I wanted.

The HN headline I commented on was "Turns out Google made up an elaborate story about me". Someone (admin?) then added "AI Overview". My comment doesn't apply to the current headline. I got what I wanted.

Yes, and I replied when it was still the original headline. I think they’re both appropriate headlines.

"Volkswagen faked emissions tests." "How dare you besmirch the good name of Volkswagen. This evil Volkswagen Passat faked the emissions tests."

Companies are responsible for the bad things they make; the things themselves are, by definition, blameless.

The original was perfectly fine. This is Google's _fault_.