Hacker News new | ask | show | jobs
by cbluth 290 days ago
I know someone that has "resurrected" a dead relative through an llm, and I've seen the nonsense on forums about dating an "ai boyfriend"...

Some people can't help themselves and don't use these tools appropriately.

3 comments

My friend admitted to having political and religious arguments with chatgpt. They have mental issues which contributed.
What's wrong with using an LLM to learn about politics and religion?

I've found Claude to be an excellent tool to facilitate introspective psychoanalysis. Unlike most human therapists I've worked with, Claude will call me on my shit and won't be talked into agreeing with my neurotic fantasies (if prompted correctly).

Because unlike a human who can identify that some lines of reasoning are flawed or unhealthy, an LLM will very happily be a self-contained echo chamber that will write whatever you want with some nudging.

It can drive people further and further into their own personal delusions or mental health problems.

You may think it's being critical of you, but it's not. It's ultimately interacting you on your terms, saying what you want to hear when you want to hear it. That's not how therapy works.

> You may think it's being critical of you, but it's not. It's ultimately interacting you on your terms, saying what you want to hear when you want to hear it.

That's been my experience with human therapists. When I tell Claude to stop being sycophantic, it complies. When I tell a human to stop being sycophantic, they get defensive.

I agree that an ideal human therapists would be better than Claude, but most that I've worked with are far from ideal. Most are not very bright, easily manipulated, and quick to defensiveness when questioned. And Claude won't try to get me to take random meds with the only justification for the specific medication being 'got to start somewhere'.

No you are gaslighting yourself without recognizing it. That's what we are talking about.
It's impossible to gaslight yourself as it requires intention.
He wants to be called a good boy, so the LLM calls him a good boy. Since the LLM is a machine that does what you want, he's essentially doing it to himself. It might not be a conscious choice, but there's still intention behind it. Kein Herr im eigenen Haus. (No master in one's own house.) - Sigmund Freud. He was wrong about a lot of stuff but this is one thing that still stands.

It's called unconscious intention, and here's a pretty interesting paper that'll bring you up to speed: https://irl.umsl.edu/cgi/viewcontent.cgi?article=1206&contex...

I'm sure you can be unconsciously intent on things, but gaslighting is a unique concept. Here's the definition I am relying on: manipulate (someone) using psychological methods into questioning their own sanity or powers of reasoning.

In your provided example, the user is obviously not trying to manipulate someone into questioning their sanity, nor power of reasoning. Quite the opposite. Lying to themselves (your example) for sure.

>My friend admitted to having political and religious arguments with chatgpt. They have mental issues which contributed.

To be fair these are probably the same people who would have been having these conversations with squirrels in the park otherwise.

We don't actually know that. It's a nice stereotype, but there's no source to prove these people would've done it anyway.
Can you talk more about the resurrection? Did they train an LLM fine tune against as much written content as possible made by that person?
That might actually be interesting if there were enough content, something of the "Beta" level AI's in Alastair Reynolds' revelation space books.

But that isn't what I've seen done when people said they did that. Instead they just told ChatGPT a bit about the person and asked it to playact. The result was nothing like the person-- just the same pathetic ChatGPT persona, but in their confusion, grief, and vulnerability they thought it was a recreation of the deceased person.

A particularly shocking and public example is the Jim Acosta interview of the simulacra of a parkland shooting victim.

the good news is that this shit isn't sustainable. when investor funtime subsidies end, ain't nobody spending $2000/mo for an AI boyfriend.

"but it's getting cheaper and cheaper to run inference" they say. To which I say:

ok bud sure thing, this isn't a technology governed by Moore. We'll see.

Mental health issues in the population are never going away, people using software tools to prey on those with issues is never going away. Arguably the entire consumer software industry preys on addiction and poor impulse control already.
yeah, I know, I'm addicted to this hellhole of a site. I hate it but I still open it every five minutes.

But that's not what I'm talking about, I'm talking specifically about people who've made a SoTA model their buddy, like the people who were sad when 4o disappeared. Users of character.ai. That sort of thing. It's going to get very, very expensive and provides very little value. People are struggling with rent. These services won't be able to survive, I hope, purely through causing psychosis in vulnerable people.

> ain't nobody spending $2000/mo for an AI boyfriend

you haven't met the Whales (big spenders) in the loneliness-epidemic industry (e.g. OnlyFans and the like)

why would they stop paying for the attention of a real woman when the artificial alternative costs just as much?
Reliability and consistency? Humans have bad days. Humans have needs and can't be available 24/7/365 for years on end. OF creators burn out or grow up or lose interest or cash out and retire.

It's not like the "real women" of OnlyFans are consistently real, or women. And there's some percentage that are already AI-by-proxy. There's definitely opportunity for someone to just skip the middleman.

Both are fake. Do you think OnlyFans models are actually giving their customers their attention?
On OF specifically, you are much more likely to be spending money talking to a bot that sends you sexy pics and messages than a real human being.
When I was at a game studio for a big MMORPG I had the valuable experience sitting next to the monetization team. It was a third grade MMO with gacha mechanics our whales spend 20-30k every month... for years.
It doesn't require a particularly powerful AI because the human's own hope is doing the heavy lifting. 70B models run juust fine on hardware you can have sitting under your desk.
People have spend much more on pig butchering scam boyfriends that don't even exist. I bet you could get some people to pay quit a lot to keep what they see as their significant other alive.
$2000? I see $4000/month minimum, roughly equivalent to what some typical Wall Street data feeds are priced at. It’s big business.