Hacker News Reader

Comment by thfuran | Hacker News Reader

You’re seriously arguing that Google’s libel shouldn’t count as libel because they showed it to too many people? It’s absolutely insane to suggest that a company should be immune from liability for its actions if it operates on such scale that those actions harm millions of people every day on the basis that dealing with that many lawsuits would be too inconvenient.

View on HN

its not inaccurate though.

consider Purdue pharma - the Sacklers got off with all their wealth intact because they were too big to sue and properly collect money for their victims.

But we agree that this was also wrong, right?

Yeah its wild to see people actually using "kill a million people and its a statistic" as a defense rather than to point out injustice.

Just because someone got away with it means everyone should get away with it ?

No they got away with it through a combination of lax pharmaceutical laws, and because they were rich and connected to the officials who should have investigated them.

They're also almost universally regarded as having committed evil acts at this point, so who cares why they got away with it?

Exactly, it's like saying Wikipedia shouldn't be liable and immediately shut down for any wrong information shown on it.

Wikipedia doesn't have to shut down, but they have to remove the libel.

The problem for LLMs is that they do not learn, and can't be prevented to produce that libel ever again. If Google finds a way to make that happen, no court would stop them from offering an LLM.

The court isn’t stopping them from offering an LLM, it’s saying that serving the output of its LLM is, in this sense, equivalent to serving the output of a human writer they hired. That is, it is Google’s speech and they are directly liable for it, unlike when they’re just serving search results that are somebody else’s speech.

Reply on news.ycombinator.com