What are you using AI for and which ones perform best at which tasks?

twopenny

That's funny Forgot 😄

Now I'll lower the tone.

Sorry I don't use it but I know people who do. Some with outstanding results on photographic and creative stuff..

But also just plain wrong. It all depends on the questions you ask and most people don't think that way. It doesn't give you the answer, just the first answer to the question. For example a friend was thrilled to get 'how do you use crutches ' but not ' what are the different ways to use crutches ' and a lot more you need to know besides.

Another wanted a photo of a stag in an autumn woodland for a show. Wrong type of stag for the background, wrong coat for the season and looked wrong altogether but he couldn't see it.

Same story, put nonsense in and you get nonsense back.

As for AI replacing staff I'm currently watching a well known website trying to do this. Unfortunately they got rid of the staff before getting it up and running, which it isn't any good at. The company is tanking fast.

jumire939

Summaries and first drafts, that's basically it. Meeting notes, status updates, and turning messy threads into something readable, everything else still needs too much correcting.

Claude's better for longer writing, ChatGPT for quick back and forth, and Perplexity if I need actual sources.

The other thing that's been useful is not a chatbot at all, Sugarbug pulls your Linear, Slack, Notion and calendar into one place so the AI actually has context instead of you pasting stuff in every time.

ZeroG82

I've been thinking about getting back into doing some coding, haven't really done any since I left unit 23 years ago. I've been told Claude is pretty useful in that department.

daveyjp

My last work report was something which a few years ago would have taken a a few days, Using Copilot and a well constructed query a report was produced in seconds. Tidying it to make it more focussed, adding some further details, graphics, photos etc then took a couple of hours.

I also used it to produce a whole load of sensitivity analysis calculations and present them in a tabular formal within seconds.

Woodstok2000

https://forums.moneysavingexpert.com/discussion/comment/82162564#Comment_82162564

This kind of stuff is brilliant for AI. I have agents managing my calendar/diary, monitoring meetings, writing minutes and summaries, producing reports etc and it saves a huge amount of time. Nothing goes out without my review, and it does occasionally make mistake but then so do I!!

victor2

https://forums.moneysavingexpert.com/discussion/comment/82162572#Comment_82162572

Human mistake(s) in that context tend to be relatively minor. AI mistakes can be the total opposite of what it should be and AI won't review it for good old common sense! 😉

Woodstok2000

https://forums.moneysavingexpert.com/discussion/comment/82162613#Comment_82162613

Thats why you always have a human reviewing the output.

victor2

https://forums.moneysavingexpert.com/discussion/comment/82162617#Comment_82162617

But then you'll get people not reviewing it as well as they should, if at all. Bit like the early days of SatNav where drivers would follow its directions literally, and end up in a field…

Vitor

You can use one AI LLM to review and critique the output of another. It's LLMs all the way down.

Woodstok2000

https://forums.moneysavingexpert.com/discussion/comment/82162939#Comment_82162939

But that's the case with any tool - it is only useful if you use it properly. People who write-off AI completely are just as bad as those who blindly use it for everything and treat the outputs as factually solid without verification.