[naillo]

I think of it as simply entertainment. It won't replace real human interaction with a partner but it's a fun consequence free environment. Tv shows act out dramatic scenes that would be traumatic and have real consequences if you experienced them in real life, but on the screen you get to play around with the idea in your head in a safe area and maybe learn something in the process.

[Fatboyrunning]

People already have unhealthy parasocial relationships with influencers.

It seems clear that people (lonely/depressed people especially) will overdose on this sort of thing once it is developed, commercialised, and less bleeding edge.

It's vapour filling the place of human connection. It's stevia. It's not going to give you cancer, but it's still unhealthy and will certainly exceed the parameters of entertainment.

[logicchains]

> People already have unhealthy parasocial relationships with influencers.

At least if they switch from parasocial relationships with influencers to parasocial relationships with open source bots they won't be financially exploited by the influencers. GPT doesn't have anything to sell us.

[michaelmior]

> they won't be financially exploited by the influencers. GPT doesn't have anything to sell us.

Except the vast majority of people aren't able to host such a bot themselves, so it seems inevitable that paid hosting services for such bots will arise. Then there's just the potential for financial exploitation at greater scale.

[Vuizur]

Decent local LLMs such as Vicuna-7B are already able to run on phones with 8 GB RAM: https://mlc.ai/mlc-llm/

Right now it's a prototype, but in the future local LLMs will be everywhere.

[bavell]

I really, really hope this is the case. Trying to be optimistic about this in spite of my skeptical nature :)

[TheLoafOfBread]

Well then catch the opportunity by the balls and start offering incels AI girlfriends written in a such a way to deradicalize them and emulate real interaction with a woman. They will subsequently get less aggressive more socialized and will also pay for it. Win-win situation.

I would call it Top W University.

[Lolaccount]

"GPT doesn't have anything to sell us"

Lol ...(worth the downvotes).

Just you wait and see.

[rvense]

> GPT doesn't have anything to sell us.

It will either find something to sell us, or we will be giving it ourselves to sell.

"If you're not paying, you're the product, not the customer"

[solsane]

Not so sure about that. Wendy's drivethrough chatbot is programmed to upsell. Even the base GPT is heavily tuned to cater to a corporate audience.

[kaycey2022]

Is the upwelling behaviour introduced using prompt injection into a generic model or what?

[Majestic121]

The commercial company exploiting ChatGPT and making it easy to use via whatsapp integrations, video, etc... surely will

[nannal]

I believe this was the issue with replika, which encouraged people to develop emotional attachments with their 'AI partner' and then first put romantic chat responses behind a pay wall before removing them entirely a year or so later.

From the outside this could be seen as a good thing, but for someone involved in the relationship, someone who may struggle with a traditional relationship and may see this as the only available option, I'm lead to understand the event was remarkably traumatic.

[2342342343636]

> It seems clear that people (lonely/depressed people especially) will overdose on this sort of thing once it is developed, commercialised, and less bleeding edge.

Sounds much less exploitative and unhealthy than the streamer/influencer parasocial relationships these people are probably currently invested in.

[sethd]

> It's stevia.

Go on…

[flqn]

Empty sweetness, I'd guess? Tastes nice but it's not real sugar?

[TeMPOraL]

A foul tasting alternative for people duped into believing that regular artificial sweetener will give you cancer.

[EniasCailliau]

My idea too. I hope people can use it to cure their loneliness in the short term and train their social skills in a safe environment so they are more confident.

[throwuwu]

It won’t work for training social skills unless the agent has self interest and competing priorities. These chat bots have been optimized for helpful assistance. Interaction with that kind of compliance on the other side will probably make your social skills much worse.

[Gare]

User experiments with early Bing chatbot (driven by some version of GPT?) have shown that AI can be both hostile to the user and protective of their own "interests".

[TeMPOraL]

ChatGPT models can be all that and more, if you simply prompt them to be like this. You can make it simulate "self-interest and competing priorities" with the right system prompt.

This makes me think if ChatGPT-based bots could be useful for teaching social skills in a therapeutic context. A therapist could use such LLM to synthesize examples of realistic dialogue by describing people and the situation, and discuss them with the patient. They could set up a bunch of bots with (hidden) priorities and goals, and have the patient navigate a conversation/situation with them - whether as a short exercise, or a prolonged one (e.g. couple weeks of talking with ChatGPT-powered fake "friends" on an IM app).

In fact, take the last bit as a free startup idea: a platform for psychologist and therapists to set up GPT-4 (or whatever comes next)-powered chatbots, with an easy interface for configuring their personalities and setting up scenarios for patients to navigate, and helping evaluate their progress over time.

[hammyhavoc]

A consequence-free environment doesn't necessarily mean a lack of consequences psychologically though.

[naillo]

Same with games already though

[hammyhavoc]

Precisely. There is healthy usage of anything.