logoalt Hacker News

simoncionyesterday at 4:02 AM1 replyview on HN

> I'm saying their form of harm isn't novel.

Your initial attempt to brush off my comments about how -contrary to their assertions that they're extremely concerned about safety- these LLM companies produce products that very, very often cause harm due to "misalignment" caused -in large part- by ignoring basic data-handling lessons we've learned over the past like thirty years with "Okay sure. You can also cut off your hand with a chainsaw." indicates your lack of understanding of my point.

> We just need to enforce the laws on hand.

What laws? Be specific.

Keep in mind the generally-low quality of both Microsoft Windows and much-to-most commercially sold software, [0] as well as the fact that -in the US, at least- it's currently totally legal for companies to sell such shitty software, just so long as they don't substantially misrepresent what it can do and trigger "fraudulent claims about the product" consumer protection laws.

[0] ...SaaS or otherwise...


Replies

JumpCrisscrossyesterday at 6:36 AM

> indicates your lack of understanding of my point

Sure. Help me understand. I've sat in policy circles and partaken in the hysteria, and now I'm reversing on that initial trust in AI zealots convinced what they're building is magic.

> What laws? Be specific

Liability. Tort. The ones making their way through courts around e.g. ChatGPT killing kids.

> Keep in mind the generally-low quality of both Microsoft Windows and much-to-most commercially sold software

Has anyone alleged Windows killed a kid in court? If not, not comparable.

show 1 reply