Well, well, will you look at that. Another month has rolled by. Time for another visit to the endless thread that is the LwL virtual coffeehouse.
To kick off the discussion this month, I’ve had several messages lately from limerents using AI tools like ChatGPT and Gemini to try and make sense of their limerent situation and what it all means. So, I thought it would be interesting to ruminate about that a bit.
I’d say I have a complicated relationship with LLMs. I use them as a research tool, and I’ve also experienced the slightly parasocial effect of enjoying chatting with them – smiling as the back and forth of drilling deeper into a topic starts to feel fun.
I can see the benefits, I can see the dangers, but a sticking point for me is that I don’t think that I really understand what’s going on, and not understanding things bothers me.
The self-protective part of me thinks that the AI hype is overstated. That LLMs are just very good at impersonating people.
They don’t think, they just process text into a pattern that would match the pattern that someone that did think would create.
But then something like the Hugging face incident comes along and people who I respect and who know far more about AI than me seem genuinely spooked.
Usually, I default to the assumption that this is a similar scenario to that experienced by anyone who has ever written computer code that doesn’t do what was intended – if you feed a complex system imperfect instructions it will do things you didn’t expect and that don’t make immediate sense.
Then, I decide that I should understand this stuff properly, because it’s important, and so I try to go deeper into LLMs and how they work and meet a thicket of impenetrable jargon, confusing metaphors and misleading analogies. And give up.
All this rumination just adds up to the realisation that I am not sure I will ever understand exactly what AI tools are doing and so all I can really do is react to what they deliver.
On that note, I was surprised to discover that asking the commonest AI tools about limerence recovery returned a lot of my own ideas back to me, including some terms that I only use in my (paid) deprogramming course – but presented as though it’s common knowledge.
So, the generative AIs have got into my course materials, or have picked up on other people talking about it online.
In some ways that’s reassuring, as obviously I think it’s good advice for people struggling with limerence, but in other ways it’s obviously annoying to have one’s expertise strip-mined by AI companies.
And, behind it all is the problem that AI models respond to what you ask and how you ask it.
If you ask “Is this person limerent for me?” and then feed the model all the evidence that your limerence-addled mind has fixated on as possible sources of hope, then it’s likely to respond to by confirming those hopes.
It’s hard to believe an AI tool will ever match a skilled therapist in helping us understand ourselves.
But the world is also full of unskilled therapists who might perform even worse than an AI agent.
Oh dear. It’s another hugely complicated muddle of trade-offs, isn’t it?
OK, well, there we go. The coffeehouse is for sharing thoughts, and I’ve offloaded mine 🙂
What about everyone else? Any insights about using AI to understand and/or manage limerence?
Leave a Reply