ChatGPT BEtter than human?

[thatsthatsaidthemayor]

I’ve had a lot of difficulties. I’ve been through 4/5 counsellors over the years. Some have helped more than others. I love my current counsellor. But , (everything before but is bullshit but not here) I’ve started taking my problems to ChatGPT and finding the answers mind blowing! If this was a person I’d be delighted but feel uncomfortable that a computer is giving me better advice on being a human than a human. I get the wealth of k owl edge that it has, and that has come from humans. It’s just weird? No? So AIBU to take advice from a computer?

I'd say, with the hallucination rate lowering substantially every day, and it not having bias, I'd trust it more than a Mumsnet post tbh.

Therapy isn't some mystical secret, it’s often just having your own thoughts reflected back to you in a way that helps you spot patterns.
If someone uses AI to realise they’re stuck in a certain cycle or to find a new way to frame a problem, that is a therapeutic insight. The 'insight' happens in the human brain, not the computer. Ridiculing people for finding a tool that works for them feels a bit unnecessary.
If it helps someone reframe a problem or feel less stuck, the benefit is real to them.
It’s okay to let people use tools that work for them.

Except that the OP talks about “taking advice” and “a computer is giving me better advice on being a human than a human”, and there are constantly recurring threads about it being “better” than a therapist. Now if it was just “I like to talk at the computer to see what it comes up with” then fine. But treating it as therapy or “advice” is deeply unwise. There’s no human safeguarding or oversight, just a mixture of the LLM output and a psychological investment in believing it’s true. Humans are wired to think anthropomorphically; and believing that an LLM has “insight” relies on exactly the same projective psychological mechanisms that have us giving cars an imagined personality, or thinking that a personality test is really insightful about us.

Just a few posts up from here, a pp talks about “giving it an entire WhatsApp conversation between me and a relative who is very hot and cold with me. it very accurately gave me insights that I hadn’t seen, patterns in communication that I hadn’t noticed and which now I see are completely right.” But who says these insights are actually right? It sounds like pure confirmation bias and wishful thinking. The LLM has no “insight” at all. It feeds back plausible text patterns to the user, and it’s the user who projects the illusion of “insight”.

“It not having bias” is a total delusion. Of course it contains all sorts of “biases”.

But proper therapy is precisely not “having your thoughts reflected back to you”. That’s mere pop psychology, not therapy - and no professionally trained therapist would ever think that’s just what therapy is. It isn’t “reframing a problem” or “helping someone feel less stuck”. That’s chit chat, and self-help, not therapy.

Actually, 'reframing' and 'reflective listening' are foundational elements of many evidence-based practices, including CBT and Rogerian therapy. While a professional relationship offers a depth a tool cannot, dismissing these techniques as 'mere chit-chat' overlooks how cognitive change actually happens. It’s about meeting people where they are.

In terms of Mumsnet bias. Yes, it has systemic bias. I meant more in the 'as a mum I xyz', 'as a mother in law I think xyz' should have been explicit about that.

People believe others on Mumsnet without researching beyond. Often agreeing only with those that reinforce their own thoughts. More recent models push back and are not sycophant chatbots. Mumsnet can be an echo chamber.

The point is, there is no need to ridicule those who choose to use LLMs in the way OP has suggested. The OP is not being harmed by talking to Chatgpt.

I have a meeting this afternoon I am very anxious about. I talked to chat GPT about it this morning. It gave me ways to help to regulate my nervous system today in preparation for the meeting. Suggestions about what to eat, what to avoid. Prompt suggestions for meditations which would help me today and types of meditations to avoid today. It gave me some regulation techniques to use a hour before, 5 minutes before and during the meeting to help me.
I helped me organise my thoughts, helped me prepare questions and answers for the meeting.
This has been increadibly helpful for me today. I am much more regulated and prepared for this meeting. I am very grateful I have this tool to help me.

I find it more helpful for parenting and relationship advice than wading through a mumsnet post.

The insights are based on huge quantities of learning from therapeutic texts online, discussion forums, research papers etc - ChatGPT 5 has the 'intelligence' level of a PHD student. It's just as likely to give a genuine insight as a human is at this stage, it just doesn't have insight itself into what it's saying because it's not conscious. That doesn't mean the information it synthesises from its training data isn't useful or meaningful.

You could have got those from Google and distilled the answers, but it would take longer. Chat GPT isn't at a different level, it's just faster, and uses more water.

Or I could have read 5 or 6 books which would have been my preferred option.

The real value isn't just speed, it's the back-and-forth dialogue that helps me think through a problem, something a static Google search result just just cant do. Google gives you data, but it doesn't help you 'think through' the nuances of a specific question in real-time. That is what helped me this morning and many mornings.

But likely would have used the same amount of water searching Google for an hour to get half of what ChatGPT was able to give me in a couple of minutes.

She could have got the answers by going to a library and taking out the relevant books! It's a new tool that exists and it's better and easier than using google in many circumstances. Why are you advocating for people not to use a tool that makes things better?

It doesn't make things better - it makes things faster and it makes things appear like a conversation.
Those are different things. It isn't a conversation because it isn't a human, and faster isn't always better.

I think you're getting a bit bogged down in the semantics of what a 'conversation' is. Whether it’s a human or an algorithm on the other side is secondary to the fact that it actually worked. If ChatGPT helps someone regulate their nervous system and go into a meeting feeling regulated and prepared, then it’s certainly helping them. It’s a bit odd to prioritize the 'human' label over someone’s actual mental wellbeing.

I honestly think if you really need a computer tool to tell you strategies to relax before a stressful day, then have at it. It isn’t having a conversation with you, though.

What it isn’t is anything beyond the level of the average self help magazine article. The reduction of therapy to “reframing” upthread is reductive and misguided. “Reframing” as one small part of a therapy is not remotely the same as “looking at an AI answer and saying that’s reframing”. It isn’t. It might sound a bit like it, but the essential problem with AI is a belief that something that is a bit like something else is the same as something else. (CBT is quite a limited form of therapy in any case, but even so, it isn’t the same as typing something into ChatGPT, even if you think you’re using the same word “reframing” to describe it.)

Saying it has the “intelligence of a PhD student” is rather amusing. The overhyping of these tools is quite OTT! So many people on this thread who are very very invested in projecting all sorts of ideas onto AI! Maybe you need to “reframe” why you are so invested in it? Why all the fervent hype around something that is explicitly designed to produce essentially mediocre output?

But “feeling more prepared” isn’t therapy, and neither is genetic “mental health”. Calling it “I got some strategies from Chat GPT on how to feel more prepared for a meeting” sounds perfectly normal. You could substitute X self-help book or magazine instead. What the AI is essentially doing is synthesising similar matching outputs from loads of self-help style material very fast. But the actual content is no more “intelligent” or personally tailored than if you’d just read it in Dr Raj or Michael Mosley or Psychologies magazine.

All that is not therapy, nor is it therapeutic insight. It’s just very fast self help style pop psychology. Great! For what is is. What it isn’t is anything mystical or anything approaching genuine insight into one’s own life.

Let’s reframe “getting a bit bogged down with the semantics of” as “random thing that’s a little bit like another thing”. Calling it a “conversation” matters if it really isn’t one. The entire point of AI is that it’s meant to look like interaction when it isn’t. Something that looks a bit like a conversation but is missing all the key features of a conversation isn’t actually a conversation.

AI is better than humans at some things and not others. It collates unimaginably large quantities of information at an eye-watering speed that no human can match.
It creates basic designs in about 5 minutes - I use it a lot for designing Facebook posters for local events. A job that used to take me ages is done in a flash.

When it comes to counselling I am less sure about it. There have been well documented incidents where it has agreed with what the user says to the detriment of that person. But I think that adaptations have now been made to avoid that scenario.

I understand that you can state specifically that you are not simply looking for an echo chamber where your every utterance is agreed with. I do use it for advice on medical resources - e.g. where can I get a cardiac procedure done the quickest. It issues very clear cautions about everything it posts. I have found that it will challenge what I ask - e.g. this is not the wisest course of action etc.

I think it depends how serious the problem you are posing is. If it is simply to say you are feeling cut up about something within say a relationship and need a different perspective or some ideas on approaching it, then I do not see major harm and possibly some benefit. If it is about serious issues that go deeper then clearly a human counsellor is the best option - but these are hard to come by on NHS and costly privately so not always easily available. A lot depends how vulnerable you are feeling and whether you are able to stand back a bit and make a judgement on the advice that AI is giving you - anyone feeling desperate may clutch at straws that may not be sound.

But, to be fair, I have seen situations in my job as social worker where crap advice has been given by counsellors.

I once needed to write a letter on a slightly delicate situation with a group I am in. I outlined the scenario, uploaded my planned letter and it took it to pieces, suggested very sensible ways of adapting it, asked questions about what my aims were so it knew what I was trying to achieve, suggested ways that my letter might be interpreted by the recipients etc. I finished up with a much better letter, but I was selective about how I used the advice.

Using AI needs some discriminatory skills and sometimes those using it for counselling are not in the sort of place emotionally where they can be discriminating.

I would say that if you are finding some help from using it and feel you are able to check the advice given against real life norms and standards and sources, then go ahead.

I am not one of those who fears AI - it is new phase in human progress and every new development has been regarded with suspicion to begin with.

*generic not genetic!

@catinateacup The reason people are 'invested' in this isn't because of hype, it's because of accessibility. Not everyone has 24/7 access to a therapist or a support network. If an AI helps someone understand a confusing social interaction or move past a moment of dysregulation, it has value. Criticising people for finding a tool that works for them is unnecessary and unkind. For many, these tools provide a bridge to self regulation and mental clarity that wasn't there before. Maybe you need to reframe and consider why you’re so intent on belittling people who are just trying to help themselves.

Wait till you find out how much water it takes to make one pair of jeans.

10-50 questions from Chat GPT= 2 litres of water.
https://www.thetimes.com/article/9167a8a8-96d1-4a68-9a13-824d862f627a?shareToken=247de510629a14e5379b60b8f2098a3c

So over a year of making 20 queries a day that's 12,000

One pair of jeans about 4000 (estimates seem to vary from between 2000 and 10,000. I buy most of my jeans second hand and keep them for 5-10 years. I do have about 4 pairs though so assuming the previous person had them for 2 years that's just under 3000 per year.

Yanbu op. I love chat gpt. So long as you keep in mind that it’s a text guesser, then you’ll be fine. It’s free and I’d say it beats therapy.

The major water usage was in the set up. It is less now that it is up and running and not much different to buying online.

The article says to cool the processors as they supply answers, unless you have better information?