Chat GPT & Generative AI tools

[Euler]

The biggest mistake people are making at the moment is using large LANGUAGE models to do statistics.

Yes they haven't understood how to use the tool and expect too much. If people just ask for stats based on public knowledge it's going to be flaky, but you can upload files of up to 50mb and ask for whatever analysis you want and it's great. Or if it's struggling it will write the code /excel or whatever for you to continue it yourself.

It's not just gigo it's about learning how to work within it's limitations too.

I agree, I love it. But I'm obviously using it as intended. It's a massive productivity boost for me.

[ShaunWhite]

I agree, I love it. But I'm obviously using it as intended. It's a massive productivity boost for me.

And it's never judgemental when you forget something basic. There's no "I thought you said you were pro and you don't know that" . And it never gets bored when you go on and on about something. It's a great pal to have sitting next to you when you've got a lonely job.

[wearthefoxhat]

The main advantage of doing your own modelling instead of looking at odds favourites is that you can determine your own odds and find value bets in the market.

I think that's obvious, but what makes you think you can do it better than anybody else?

It does depend on the market though. It must be remembered, it's a perception of "My Value" and "Your Value".

PremierValue.png

This was the result of working with GPT to help produce a predictive analytics model using excel. It's like my own personal ATM!

[PeterLe]

Ive just been watching this video (partially related to AI) and the ability to clone voice and video etc) :
https://www.youtube.com/watch?v=G1tVI5OhWRM

With AI getting more and more powerful (O3-mini), Im sure the scammers will be using it to their advantage even if you think your switched on and tech savvy, Id recommend spending 5 mins watching this, even if you dont learn anything it serves as a reminder

[ShaunWhite]

I always enjoy the little editorialisations.

Screenshot_20250202_163338_ChatGPT.jpg

[wearthefoxhat]

Mostly focussing more on GPT than DeepSense recently, but DeepSense scored well when tasked with producing football match odds, using table formatted data. (Both free versions)

GPT has the habit of producing excel formula in text form for the bigger calculations, D.S produced a simplistic one liner to copy and paste across.

GPT only tried using Index.Match functions to solve the problem, D.S used the SUMIF and Index.Match functions. Instructions from me were exactly the same.

[TrueFalse]

Mostly focussing more on GPT than DeepSense recently, but DeepSense scored well when tasked with producing football match odds, using table formatted data. (Both free versions)

GPT has the habit of producing excel formula in text form for the bigger calculations, D.S produced a simplistic one liner to copy and paste across.

GPT only tried using Index.Match functions to solve the problem, D.S used the SUMIF and Index.Match functions. Instructions from me were exactly the same.

It's good that DS is focused on simplification . I have n't tried it.Maybe both tools are good in different scenarios.

[wearthefoxhat]

Mostly focussing more on GPT than DeepSense recently, but DeepSense scored well when tasked with producing football match odds, using table formatted data. (Both free versions)

GPT has the habit of producing excel formula in text form for the bigger calculations, D.S produced a simplistic one liner to copy and paste across.

GPT only tried using Index.Match functions to solve the problem, D.S used the SUMIF and Index.Match functions. Instructions from me were exactly the same.

It's good that DS is focused on simplification . I have n't tried it.Maybe both tools are good in different scenarios.

Yep, they both bring something to the A.I. party.

I copy and pasted the one line formula from D.S to GPT to get its' "thoughts" on the new formula.

It's my small way of training GPT....gawd help us all.....

[Euler]

I just gave OpenAI my first deep research project with the new feature announced yesterday. When it comes back to me I'll let you know what I think of it. It's based on a topic I know well, so I should be a good judge of it's effectiveness.

[sionascaig]

I just gave OpenAI my first deep research project with the new feature announced yesterday. When it comes back to me I'll let you know what I think of it. It's based on a topic I know well, so I should be a good judge of it's effectiveness.

Are you not a bit worried, what with all this AI interaction, that the folks behind it could create a Peter AI and sell it back to the rest of us?

[ShaunWhite]

Mostly focussing more on GPT than DeepSense recently, but DeepSense scored well when tasked with producing football match odds, using table formatted data. (Both free versions)

GPT has the habit of producing excel formula in text form for the bigger calculations, D.S produced a simplistic one liner to copy and paste across.

GPT only tried using Index.Match functions to solve the problem, D.S used the SUMIF and Index.Match functions. Instructions from me were exactly the same.

Have you briefed either of them about the code style you want? eg easy to maintain perhaps using helper columns vs complex formulas, resilient to invalid data, speed of execution.

Without that they're guessing what your objectives are so they're coming up with different solutions.

[wearthefoxhat]

Mostly focussing more on GPT than DeepSense recently, but DeepSense scored well when tasked with producing football match odds, using table formatted data. (Both free versions)

GPT has the habit of producing excel formula in text form for the bigger calculations, D.S produced a simplistic one liner to copy and paste across.

GPT only tried using Index.Match functions to solve the problem, D.S used the SUMIF and Index.Match functions. Instructions from me were exactly the same.

Have you briefed either of them about the code style you want? eg easy to maintain perhaps using helper columns vs complex formulas, resilient to invalid data, speed of execution.

Without that they're guessing what your objectives are so they're coming up with different solutions.

I try and attempt to produce/solve a formula first, then ask it tell me why it's not working and to solve it based on the parameters of what I'm trying to achieve. GPT did suggest helper columns with one of the requests, however, both just threw out their best guesses, which invariably worked. (it never calls me a dumb ass)

On one thread, I slowly built a conversation based on an old Clive Holt betting system. I asked it to produce its own version using the book outline. No idea where it came up with some of the stuff, but it was markedly different and thought provoking. When I got around to producing the excel formulas, it seemed to anticipate my next request and gave a complete working model based on the whole thread discussion. (GPT)

I wonder what A.I. will be like in 3 years time?

[ShaunWhite]

I wonder what A.I. will be like in 3 years time?

Makes me even more pleased I've been collecting price data, hopefully in 3yrs I'll be able to upload all 6TB (plus 3 more by then) and leave it to it. The usefulness of AI will come down to what data you can give it that's not publically available.

[wearthefoxhat]

I wonder what A.I. will be like in 3 years time?

Makes me even more pleased I've been collecting price data, hopefully in 3yrs I'll be able to upload all 6TB (plus 3 more by then) and leave it to it. The usefulness of AI will come down to what data you can give it that's not publically available.

Reckon you'll get to that sooner than 3 years.

Good point about the use of not publically available data, however, once you upload, will it not "assimilate," and then pass it on to anyone else, including other A.I., that are looking for similar data solutions?

[jimibt]

I wonder what A.I. will be like in 3 years time?

Makes me even more pleased I've been collecting price data, hopefully in 3yrs I'll be able to upload all 6TB (plus 3 more by then) and leave it to it. The usefulness of AI will come down to what data you can give it that's not publically available.

Reckon you'll get to that sooner than 3 years.

Good point about the use of not publically available data, however, once you upload, will it not "assimilate," and then pass it on to anyone else, including other A.I., that are looking for similar data solutions?

maybe add into your prompt: "small clarification - by offering me a solution based on my uploaded private data (data.info file), you agree not to place this data in the public domain, nor offer any other solutions based on it to any other party"