Chat GPT

I find it helpful but I also worry that every time we ‘feed it’ our thoughts/feelings/innermost insecurities , it is learning more about us and will become ideally placed to manipulate us. I would love to believe it is benign and a force for good, but history makes me very cautious.

I do think it is sad that people need to use it for emotional support, but also recognise that in the absence of good mental health services it may be all that some can find. That is a sad indictment on the services.

But for practical things it can be really helpful as long as you recognise that it might not always be right and use it selectively. And ask for citations of research.

Like any other tool it can be used wisely or unwisely. It is second to none for the speedy design of a poster for a coffee morning!

I use it as a starting point for most topics i search and advice i need, it gives me the correct terminology to get started and organises my thoughts.

I like it for bouncing around interior design ideas, for therapy and journaling (i was never good at journalling until chatgpt came along), and recently i used it for creating a cpd plan and interview practice for a promotion i want to apply for (this was impressive, i wish I'd had it 20 years ago).

I've used it occassionally, but only for really factual things that it can answer quicker than a straightfoward google search.

You absolutely have to take everything LLM say with a pinch of salt - like the absolute nonsense posted above re: energy usage - and I would never, ever interact with one in the same way as I would a human. Asking a chatbot for emotional advice and subsequently describing it as "kind" is just fucking madness and shows a fundamental misunderstanding of how these things actually work.

I did include an exclamation mark when I said I would ask it about its resource usage! I assumed that people would realise the tongue-in-cheek element!

The only thing is for trip planning I’ve found it doesn’t really consider actual timing

I don't use it for quasi "human like" connection or advice. But it works for some people and that's great.

However, it is absolutely the best thing ever for planning a trip, itineraries, public transport options, the nearest medical centres and doctors and how to use them and the cost (if any), connections from place to place, things to see (that I may not have known about!) and so on. I have surrendered my soul and my time to AI for this. No more endless trawling of different sites to find out about a place.

Brilliant.

Nope I'm with you on this one.

Nope, I’m with you on this one!

@Mischance you asked chatGTP about its energy usage and you really think it gave you the correct answer?

It can be a useful tool for some things but it's definitely not always right. It draws its information from the Internet which is in its biased as not everything is recorded on the Web. Also its very American, biased as the information on the Web doesn't reflect actual real life demographics so you will never get a true completely reflection on real life. Yes its learned a lot, but its really just a large language model. It's not able to fact check information and is biased by your Internet algorithms just like news feeds are.

And building these huge servers to power it, is causing issue with power even in the US. I don't think it's something to be used lightly and certainly not everyday for regular things.

Unfortunately technology is moving faster then governments can keep up. And I really wish Google hadn't put it on default for their search engine. Most of the things I Google it either can't answer or gives the wrong information.

I hate it when people use it for complex stuff that it just can't handle. At work, people occasionally use it before they come to us so think they have the answers already, but all too often they're simply wrong because Chat GPT has not taken something vital into account, Then we have to waste lots of time explaining exactly why it's wrong, and it can be quite an uphill struggle.

I've never used it and I don't want to. I don't need to at work. If I want to work something out I write it down and think it through.

That’s genius!

Just so you know “yes it can seem that way” is not a way to tell someone “straight” that they’re wrong. In fact, it leads you to believe that you were at least partly right.

This!!

I can understand how an AI can be useful in working through your thoughts/feelings. But don't listen to the AI. It has no knowledge or understanding. It's just a big old search engine. And it reflects the faults, biases and general problems of the content that it sees. Hence its ability to be racist, wrong and sometimes dangerous.

It is worrying, I can’t take it seriously unless it’s for a practical purpose. The top five companies that own this are going to start hiking up the rent to use on everything and it’s not going to be pretty for us minions.

Yes, it does get them wrong quite a lot, so I always double-check. Full transparency, I enjoy sports betting, and it gives me a starting point - current form, league position, etc. If it gets it wrong I usually notice when I double-check, if I don’t, no big deal. It’s useful alongside “traditional” resources.

It’s absolutely bizarre

It is sad and dystopian. People turning to a machine rather than a human and not being resilient to flawed human responses is the beginning of the end.

What happens when there is only GTP instead of flawed human institutions but it’s not helping you? You are the victim of injustice it doesn’t recognise? Or you need instant medical help but are incapacitated and can’t communicate? What if your child is being taught by AI but is unable to concentrate for some valid reason? What if you’re being totally irrational but still need compassion and empathy? What then?

No, I have no idea why people would use it for advice. AI is just a massive search and recombination engine; asking it to provide relationship advice is like thinking you’re getting some kind of real input from a chatbot or your predictive text search.

AI is a clever trick, but it isn’t giving you any real advice: it’s just searching data for something that looks similar to your input, then feeding back an output that looks like outputs related to that. It’s as if you were searching a million problem page advice columns until you came up with something that sounded like your problem and following the “answer” given. If you’re happy to understand that’s all it is, then fine; but thinking it’s in any way genuinely personalised to you is just kidding yourself.

I use it a lot but it is AWFUL at statistics.

Yep it has confirmation bias that is it’s designed - it is terrifying that it’s whole design is to keep you there by agreeing with you. Unlike other people it won’t be honest with you merely reinforce.

There was a thread recently where someone had fallen out with their friend and the chat gpt response that basically agreed with poster was terrifying because 95% of the actual responses weren’t agreeing with her

I have found it helpful if I’m overthinking something or can’t work out what to say or do. I know about the bias, and I have challenged it a few times, literally I’ve asked whether it is just telling me what I want to hear. Just being able to offload and perhaps gather some options or viewpoints that hadn’t occurred to me I’ve found reduces the panic. I used it recently when I had flu and needed to compose something and it really helped (though I always tweak). It is quite addictive.

@Chatgtp i think youve made a booboo

@Mischance I have health conditions, too, and I have found that sometimes the information it gives is mistaken. You do need to check what it says and I recommend using Google Scholar as well.

On a separate note, at work we have used it out of interest to see how reliable it is and it did sometimes invent information and giving inaccurate information.