Quick take… A fairly obscure Canadian musician named Ashley MacIsaac lost money, and will likely lose a lot more, because Google’s explainer AI told people that he was a sex offender. Apparently there is a known sex offender with the same name.
Constants and variables show that AI itself isn’t the main factor here.
Constant:
This is a straightforward libel case. A widely trusted publisher made a clearly false statement about an innocent person, leading to loss of money and reputation.
Variable:
Before Google started using its AI explainer, Google itself couldn’t have been responsible for libel. Google acted like a phone book or a library card index, linking to information available elsewhere. If you were planning to hire this guy for a concert and you weren’t familiar with him, you might look up the name on Google. You could easily determine that the MacIsaac who played the violin was an entirely different person from the MacIsaac with a criminal record. If you then fired the violinist and spread false rumors about him, you would be committing libel.
After Google started up its explainer, Google’s “official” reports got messy and confused, without the judgment and context that a competent human would apply. Google could have used humans to create messy and confused reports, with the same result.
Conclusion:
Google gained well-deserved trust by serving as an uninterpreted card catalogue for 10 years. Now people still trust Google, but its results are no longer trustworthy.
Polistra's Mill 