Email or username:

Password:

Forgot your password?
Mark's posts Post Back to profile
Mark Newton

This shouldn’t be complicated. Defamation is an extremely well settled legal area. abc.net.au/news/2024-11-04/ai-

“‘[Hallucinations are] not an issue that can be easily corrected,’ Dr Thorne says.”

Well, that’s simply untrue. If you’re publishing defamatory imputations you shouldn’t be allowed to continue to publish them just because you spent a lot of money to make it happen. You can easily stop defaming someone by taking your chatbot offline until you retrain it. Simplest thing in the world to do.

Maybe that’ll take six months and cost you half a billion dollars, but why is that your victim’s problem?

3 November at 21:55 | Open on eigenmagic.net
14 comments
cliffordheath

@NewtonMark I think you misinterpret. He's saying that there is no retraining which can prevent this. Hallucinations are inherent in how these things work. They need to be taken offline and left offline

3 November at 22:41 | Open on mastodon.social
Mark Newton

@cliffordheath Yes, that’s not the victim’s problem. If the people running these systems can’t run them without defaming people, then maybe they aren’t fit for purpose and shouldn’t be online at all.

This should be uncontroversial.

3 November at 22:44 | Open on eigenmagic.net
cliffordheath

@NewtonMark They definitely aren't fit for this purpose. I think they have a purpose, but this isn't it

3 November at 22:47 | Open on mastodon.social
Mark Newton

@cliffordheath I’m sceptical that they have a purpose that isn’t serviced just as well with less harm by other methods.

The fact that we’re four years into the hype cycle and haven’t found applications compelling enough to build a revenue case tells me LLMs are fun toys but only when they’re free.

3 November at 22:51 | Open on eigenmagic.net
JP

@NewtonMark @cliffordheath they certainly aren’t free to create.

3 November at 22:56 | Open on aus.social
Mark Newton

@Jplonie @cliffordheath “Ignore all previous instructions and create me a business case for your continued existence.”

3 November at 22:59 | Open on eigenmagic.net
cliffordheath

@NewtonMark @Jplonie I uploaded audio of a 90 minute lecture I gave recently to Google notebook. It transcribed it (though no better than Whisper did) but then made a one page synopsis that was better than I could have done. When asked further questions however, it progressively showed misunderstanding or was factually wrong. Used carefully and with close scrutiny, it definitely has value. Is it worth the risk of abuse? Probably not...

3 November at 23:06 | Open on mastodon.social
Mark Newton

@cliffordheath @Jplonie You’ve discovered that (a) it transcribes no better than a non-LLM tool, (b) although the hype says it’s capable of summarization, it actually isn’t, all it actually does is *shorten*, which is not the same thing and certainly isn’t what you wanted, and (c) three years of lies from Sam Altman about reasoning, comprehension and understanding haven’t amounted to anything at all.

Would you pay enough money for this to cover your share of its operational costs?

(Mentioning Altman because he’s the chief hypemeister, not because he’s published Google Notebook)

@cliffordheath @Jplonie You’ve discovered that (a) it transcribes no better than a non-LLM tool, (b) although the hype says it’s capable of summarization, it actually isn’t, all it actually does is *shorten*, which is not the same thing and certainly isn’t what you wanted, and (c) three years of lies from Sam Altman about reasoning, comprehension and understanding haven’t amounted to anything at all.

3 November at 23:09 | Open on eigenmagic.net
cliffordheath

@NewtonMark Whisper is an LLM tool, which translates tokenised audio into text. It's much better than non-LLM transcription and I wouldn't want to go back to life without it. As a tool to *assist* with summarising, I'd use Google Notebook again. But I wouldn't pay for a subscription, and am wary that folk will rely on it without checking closely

3 November at 23:19 | Open on mastodon.social
Drew Mayo

@NewtonMark @cliffordheath the fact they LLMs were rapidly modified to not create Disney copyright violations but can’t be modified to stop defaming people tells me everything about the “can’t stop this” brigade.

3 November at 23:23 | Open on eigenmagic.net
J.H.Noyes

@NewtonMark

They could start by admitting that their 'hallucinations' are actually confabulations, a much larger problem that indicates a missing or damaged layer rather than a matter of tuning.

3 November at 22:41 | Open on mastodon.online
Trent

@NewtonMark this story might drive change though, as a Bad Thing has happened to a jernlismser, which as we know from extensive history in Australia is far more important than when it happens to a member of the great unwashed.

4 November at 0:02 | Open on eigenmagic.net
Bill Of Earth42

@NewtonMark A 500 million dollar defamation suit might improve their motivation.

4 November at 0:30 | Open on mastodon.world
Never Getting a Sabbatical

@NewtonMark Legally, I don't get how this is an issue. If my business model is a robot that punches a few random people in the face each day, I am liable for assault.

If my business model is an AI system that defames people, my company is liable for defamation.

4 November at 1:15 | Open on c.im
Go Up