Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

AI: Markets for Lemons, and the Great Logging Off

fortressofdoors.com

Post reply on HN
Print view View original

181–190 of 292 posts

Re: AI: Markets for Lemons, and the Great Logging Off

#181

Post by bilbo0s »

ClumsyPilot wrote: post #148
angarg12 wrote: post #25

Regarding the Third point, I can already see part of this movement. I've become disenchanted with the quality of the discourse. Specifically as I've entered a more senior stage of my career, I'm looking to mingle with like-minded folks in similar situations. Thing is, these folks don't usually lurk online, and even if they do, their voices get overwhelmed by the flood of (mostly low quality) content. I'd be happy to…

> I'd be happy to pay a fee to become a member of a community with vetted participants. FreeMasons and premium golf clubs!

Problem is, both can still be filled with less than bright people.

If you have an interest in, say, the math behind various AI techniques. Or maybe an interest in digging into the differences between different, (but seemingly similar), cancer therapies. (Think TIL v CAR-T). You won't get that at the golf club. You will need some kind of vetted discussion group.

Re: AI: Markets for Lemons, and the Great Logging Off

#182

Post by deadpannini »

hasbot wrote: post #171

People go to social media sites for engagement. If AI bots provide good interaction, most people will be happy and won't care if they are chatting with a bot. Hell, the best social site would be one in which I could create a couple of AI companions with whom I could chat about things that interest me that day. The companions would scour the Internet for related news and topics, bring them to my attention, and we'd ch…

Reading this comment (which is correct, I think) gives me the urgent desire to log off HN.

Re: AI: Markets for Lemons, and the Great Logging Off

#183

Post by treis »

awestroke wrote: post #159
anselm wrote: post #156

Authors making this argument (which we see often now) really need to sketch out why cryptographic solutions will fail. The entire thesis rests on this erroneous sentence in the second point: “This can be done in two ways – just move to invite-only silos where you already know everybody, or big platforms where the owners do the vetting for you.” There are more options. We employ them when we are forced to; when it bec…

So, people start signing their AI-generated messages. What have we gained?

You can ban them and they'll stay banned. Today they'll come back or switch to one of the other dozen accounts they post crap on. The counter argument to that is that places like Facebook are mostly not pseudonymous and a lot of crap is posted there.

Re: AI: Markets for Lemons, and the Great Logging Off

#184

Post by peteradio »

anselm wrote: post #156

Authors making this argument (which we see often now) really need to sketch out why cryptographic solutions will fail. The entire thesis rests on this erroneous sentence in the second point: “This can be done in two ways – just move to invite-only silos where you already know everybody, or big platforms where the owners do the vetting for you.” There are more options. We employ them when we are forced to; when it bec…

You can employ those techniques but real people will get blocked as spam. And the better AI can evade the closer to the bone you will have to cut. Then what? AI is interacting with your algorithm to silence voices.

Re: AI: Markets for Lemons, and the Great Logging Off

#185

Post by tapanjk »

jnwatson wrote: post #123
tapanjk wrote: post #103

> but even if that hits us tomorrow, the amount of raw power that's been unlocked in the last year alone – the tech that's already here – already has the potential to change things in massive waves. If this is about ChatGPT, we should remember that it is not available in a way that you can build your business applications on top of it in any reliable or meaningful way. It is a very powerful demo of what AI can do, bu…

Still, it will be replicated, then optimized, and eventually cost pennies to run. The writing is on the wall.

Well, guess I was wrong. There is a product in the making. [1]

[1]: https://twitter.com/michael_vandi/status/1607422866416619520

Re: AI: Markets for Lemons, and the Great Logging Off

#186

Post by vlovich123 »

anselm wrote: post #156

Authors making this argument (which we see often now) really need to sketch out why cryptographic solutions will fail. The entire thesis rests on this erroneous sentence in the second point: “This can be done in two ways – just move to invite-only silos where you already know everybody, or big platforms where the owners do the vetting for you.” There are more options. We employ them when we are forced to; when it bec…

Because using technology to solve human problems rarely/never works. [1] was originally written for spam and bore mostly correct. Explain how replacing spam with “AI-generated spam” changes anything? You can try to fight this stuff but it will look more like AI to detect AI (similar to our current anti-spam tech). No reason to believe cryptography has some kind of magical bullet here as it’s an unrelated problem domain. And to the person claiming you get kicked off and that prevents you from coming back ignores a) we haven’t solved being able to tie disparate online personas for a unique offline one (despite Facebook ostensibly trying really hard) b) there are all sorts of secondary problems that pop up when you try to do that (eg ignores the concept of learning from your mistakes and redemption, key things that happen frequently with the young or anyone else testing boundaries).

[1] https://trog.qgl.org/20081217/the-why-your-anti-spam-idea-wo...

Re: AI: Markets for Lemons, and the Great Logging Off

#188

Post by djmips »

treis wrote: post #183
awestroke wrote: post #159

Earlier quoted context omitted.

So, people start signing their AI-generated messages. What have we gained?

You can ban them and they'll stay banned. Today they'll come back or switch to one of the other dozen accounts they post crap on. The counter argument to that is that places like Facebook are mostly not pseudonymous and a lot of crap is posted there.

And when you get banned because someone stole your key - that's going to be awesome.

Re: AI: Markets for Lemons, and the Great Logging Off

#189

Post by pavel_lishin »

> A Market for Lemons

> In which the internet gets clogged with piles of semi-intelligent spam, breaking the default assumption that the "person" you're talking to is human.

Neal Stephenson's novel Anathem briefly mentions something like this as part of the novel's worldbuilding.

Re: AI: Markets for Lemons, and the Great Logging Off

#190

Post by Jiro »

ShredKazoo wrote: post #69
Jiro wrote: post #45

Earlier quoted context omitted.

The AI could comment as well as a human until it gets to its payload, which is a scam. People will vote it down after they get taken in by the scam, which is too late. Or maybe not even then, if the payload isn't a scam per se, but something that's deliberately designed by the AI's developer to be slanted. A million AIs that are indistinguishable from humans but occasionally throw in a comment about how they got sick…

>People will vote it down after they get taken in by the scam, which is too late. Not too late to create training data by which the social media platform can detect and remove subsequent scams. Especially if they 'report' rather than downvote. >A million AIs that are indistinguishable from humans but occasionally throw in a comment about how they got sick from Pepsi could make things very profitable for the Coca-Cola…

>Not too late to create training data by which the social media platform can detect and remove subsequent scams.

How is that going to happen? We're assuming the AI posts exactly like a human until the comment about Pepsi comes out. It'll be impossible to distinguish from a human, by assumption. Or are you going to just have the site reject everything which mentions Pepsi negatively?

>How is this different from Coca-Cola paying influencers to post negative stuff about Pepsi?

AIs don't need to be paid.

Post reply on HN