A more facetious way to put "don't use LLMs": share your input prompt. The people who don't want to read LLM output won't have to see it, and those who do can enjoy using an LLM the same way the author would've. This actually saves a loss/noise step as the consumers won't need to deploy an llm to summarize the authors llm expanded output.

$ chatgpt make this comment more glib

Rhetoric machine successfully practices oratory. More news at 11.

It's good to see this studied, but this should really be more obvious.

We should always study what we think obvious, because we are often wrong.

I'm glad to see this studied. I'm just surprised at the reactions I'm seeing.

Musicians are paid to record music. If they were not, as is the case for ai data, we would indeed consider it theft.

> Musicians are paid to record music. If they were not, as is the case for ai data, we would indeed consider it theft.

I remember reading literature from the 1930s, and there were quite a few folks who thought that the musicians-doing-recordings were stealing from the old-timers who played for live audiences.

History does not exactly repeat itself, but it rhymes

The few who get paid to record are replacing far more who would have otherwise been paid to play if that was the only way to listen to any music. Is it okay if one musician replaces another but not if a non-musician does so? The person getting replaced didn't get paid either way.

Going back to the previous example, say I pay a different coworker $50 for the data to train the chimpanzee and then use it to replace the first person. In either case they lost their jobs while receiving nothing for it. In either case, what happened to them is the same, so how would they be stolen from in one case and not in another?

Whether it is okay or not has no bearing on it being theft. And no need to create a hypothetical, this is exactly what happened to musicians.

There is broad evidence that labs have used substantial amounts of pirated data, no need to reach for a new definition of theft.

Private healthcare is much worse, seemingly they have an open access policy. New breaches occur in the order of millions per week. Not remotely newsworthy anymore. (last time this was mainstream worthy was 200M leaked records in 2024). Last week https://www.securityweek.com/4-1-million-impacted-by-adapthe... Week before that https://www.yahoo.com/news/us/articles/more-9-5-million-pati... 2 weeks before that: https://www.msn.com/en-us/health/general/carecloud-confirms-...

And it will remain this was as long as the consequences of not protecting our data remain trivial.

if only we prosecuted corporations as people instead of just giving them the civil liberties of one

In my experience, any industry following a security standard halts all effort at security once they're compliant with the standard. HIPAA sets a minimum but seems to also guarantee you will get exactly that minimum and nothing better.

The only thing HIPAA guarantees is that when your data is handed out, there was a policy around it.

Not true, it makes investigating medical fraud almost impossible.

You wrote "apt apt apt apt apt apt" to prove a very specific point no? Absolutely requires reasoning and understanding the problem to go there. And good luck getting an LLM to do that.

Yes, I wrote it to prove that writing doesn't mean I understand the world. It responds to "writing helps you understand the world." My point was that, yes, of course you can find categories within it that apply, which is why the answer is not fundamentally incorrect. Claiming "writing = thinking" in a general sense is very broad, and that's my point.

> And good luck getting an LLM to do that.

Have you never asked a decent model to explain something to you? You should try it.