Live data from Hacker News

Social Skill Training with Large Language Models

arxiv.org

21–30 of 105 posts

Re: Social Skill Training with Large Language Models

#21
> However, practice environments for social skills are typically out of reach for most people.

I learned my social skills in school, interacting with my peers in my neighborhood, and from my family. I understand this is not available to everyone, but anyone with access the LLMs there is a high likelihood they also (should?) have access to schooling, some neighbors, and family.

To what culture this LLM was trained to?

Re: Social Skill Training with Large Language Models

#22
> However, practice environments for social skills are typically out of reach for most people.

This is exactly what I predicted...AI and technology isolates us from each other by destroying communities, and is slowly replacing real human interaction with computer interaction. The end result is that in the future, people will have fewer and fewer opportunities to interact with others and will have to be drip-fed satiations for their basic needs such as socialization, similar to how hospital patients are given IVs.

Before you say this is outrageous, actually take a look at history and how technology has pushed as farther and farther apart. Nowadays there is very little real interaction beyond the nuclear family. Think of self-checkouts, chat-bots, AI partners and therapists, technology making us more self-reliant (which is a bad thing).

If we keep letting technology take away our communities this way in this two step process (first, take away the community then supply an inferior, superficial substitute), then we will compelely lose our humanity.

Re: Social Skill Training with Large Language Models

#24

Use these models to enable people to be tolerant and value other communication styles... not to train everyone to communicate in identical fashions. We seem to be working hard to homogenise humanity. Soon we'll all be trained by the same model to generate the same answers in the same way to the same beat. The era of the McHuman. We seem to think we can use LLMs as an edge to make us richer or better than other humans…

I agree. In my own experience the vast majority of miscommunications are a matter of style, opinion, or assumption of essential facts in a domain unfamiliar to the listener.

People don't like to feel dumb. If you overexplain or underexplain you will trigger these feelings.

People also don't like confronting opinions that differ. Many followers of a given subject readily buy into the popular opinions within that space and ignore hygiene during discussion.

People also don't like reading or listening to disorganized or long-winded thoughts.

Re: Social Skill Training with Large Language Models

#25

> However, practice environments for social skills are typically out of reach for most people. This is exactly what I predicted...AI and technology isolates us from each other by destroying communities, and is slowly replacing real human interaction with computer interaction. The end result is that in the future, people will have fewer and fewer opportunities to interact with others and will have to be drip-fed satia…

Look beyond your bubble. The child participating in 'School over the Air' in Australia likely appreciates having such a resource to develop their social skills. Sometimes, you need a private space, especially if the community around you isn't on the same wavelength as you are or it's your family and some farm hands.

Re: Social Skill Training with Large Language Models

#26

> However, practice environments for social skills are typically out of reach for most people. This is exactly what I predicted...AI and technology isolates us from each other by destroying communities, and is slowly replacing real human interaction with computer interaction. The end result is that in the future, people will have fewer and fewer opportunities to interact with others and will have to be drip-fed satia…

It may take an active, conscious, participatory approach to compensate for.

Though, can you help me understand the connection between the grocery store checkout process and community a bit more?

Re: Social Skill Training with Large Language Models

#27

Use these models to enable people to be tolerant and value other communication styles... not to train everyone to communicate in identical fashions. We seem to be working hard to homogenise humanity. Soon we'll all be trained by the same model to generate the same answers in the same way to the same beat. The era of the McHuman. We seem to think we can use LLMs as an edge to make us richer or better than other humans…

But everyone wants to communicate in identical fashions... the most commended book for social skills is from the 1930s.

Re: Social Skill Training with Large Language Models

#28

One big concern I have with this is that the state of LLMs because of issues in where it can go horribly wrong, requires someone who is familiar with the subject to evaluate their output. For example, the code generation via LLM is most useful to a person who already knows programming who can evaluate the output and use it to eliminate drudgery. Previously, there was an article posted about using LLMs to generate pat…

Giving the underprivileged tools to work on themselves is a huge win for society, regardless of the fidelity of all interactions. There are a lot of bad therapists in the world, and not all of them are going to give good advice all the time. But, bad therapists can still provide some information that may be useful to the person learning. Why would we expect perfection from interacting with an LLM instead of a bad the…

I think the reason why LLMs pose a much higher risk is scale. A bad therapist will impact a limited number of patients, whereas an LLM deployed at scale will affect possibly millions of people. In matters of health and safety, the smallest error rate could have devastating consequences.

Re: Social Skill Training with Large Language Models

#29
Another episode of "what convoluted way can we use tech to 'solve' a social issue that it clearly exacerbates to try and desperately keep the blind cult-like silicon valley version of techno-optimism alive for as long as possible, while ignoring the obvious disastrous social consequences"

I'm pretty sure nobody 15 years ago would have uttered the phrase "opportunities for practising social skills are hard to come by". All social technology ignores our biological need for physical closeness and the severe negative outcomes to our brains and health when we don't have it. People who are so socially crippled that they would be considering this treatment are a symptom of destroyed communities, broken social cohesion, tech addiction and this overly popular brand of post-covid hermitism that seems popular on this site. These are social ills that all need to be tackled. Absolutely nothing about talking to an LLM will help,

Re: Social Skill Training with Large Language Models

#30

> However, practice environments for social skills are typically out of reach for most people. I learned my social skills in school, interacting with my peers in my neighborhood, and from my family. I understand this is not available to everyone, but anyone with access the LLMs there is a high likelihood they also (should?) have access to schooling, some neighbors, and family. To what culture this LLM was trained to?

Right? And even if you are in such a dire position that you literally have no oppurtunity to speak to people irl - that's bad. You need to be helped to leave that situation immediately, not given an LLM. You can literally die from never having proximity to humans.
Post reply on HN