Hacker News new | ask | show | jobs
by missedthecue 7 days ago
If companies can be held liable (in spite of very visible disclaimers, ToS, and usage policies) for the output of non-deterministic software, isn't this just a soft ban on the deployment of non-deterministic software?
7 comments

> (in spite of very visible disclaimers, ToS, and usage policies)

If you sell food, in a food stall, labeled as food and you add a disclaimer that it is toxic and will make you sick. You are still selling toxic food and you are liable for it.

Google is pretending to give answers to your questions. They offer you a service about answering questions. And then they add a disclaimer "we do not answer questions just write bullshit". That is still fraud and Google should be liable for it.

> isn't this just a soft ban on the deployment of non-deterministic software?

Tetris is non-deterministic and it is not banned like millions of other programs. I do not follow you.

To add to this, a Google search now is answering your question in an incorrect way rather than merely bringing you to a site with incorrect information on it.

They are also no longer covered by safe harbour provisions because it is them answering it, not some content they refer you to.

Where did it say the liability only applied to non-deterministic software?
If Google’s AI overviews used a deterministic LLM the ruling would hold just the same. This has nothing to do with non-determinism.

If your software deterministically produces incorrect output for some inputs, you’re liable the same as if it did it non-deterministically.

What to do if the software automatically and wrongly libels you on a public search engine?

Honestly I can understand the ruling, but the side effects might be severe.

If you read those AI overview, it is not non-deterministic, it is conclusions, claims, statements.
LLMs are deterministic, they are only non-deterministic when you add a temperature.
no, just a ban on using non-deterministic software for situations where deterministic responses are expected.
Deterministic responses are not expected from something clearly labeled able to make mistakes.
it's not clearly labeled. it's a search engine. i expect deterministic responses from that.
It literally says “AI can make mistakes.” And yeah, I’m sure everything else on the Internet was true before.
google search results were true in that they accurately reflected the content of the pages that were found. it would then be up to me to decide if i trust those actual pages.

that trust level obviously varies, but for example if the page found is from wikipedia i know that the content has been reviewed by multiple people, so i can trust that more than say a blog post on the same topic.

on the other hand, AI results are not certain to be an accurate reflection of that wikipedia content, and therefore it is less useful than the page itself.

practically speaking i don't even look at the results in detail. i just check if my search keywords match and then i open the pages and read them myself.

the AI summary does not tell me which keywords on the pages matched. it is therefore completely useless.