z036.png
Chat GPT & Generative AI tools
-
jamesedwards
- Posts: 6301
- Joined: Wed Nov 21, 2018 6:16 pm
You do not have the required permissions to view the files attached to this post.
Could have been a credible defence for the Australian poisoner
-
ShaunWhite
- Posts: 10881
- Joined: Sat Sep 03, 2016 3:42 am
Some might find this interesting...
Betfair Australia shooting the breeze about AI
https://youtu.be/gODvtuW4sjA?si=lg1j1Yxk7wm8MJPw
Betfair Australia shooting the breeze about AI
https://youtu.be/gODvtuW4sjA?si=lg1j1Yxk7wm8MJPw
I just put up a substack article on my take of AI vs more traditional statistical analysis
https://open.substack.com/pub/hoofs/p/i ... st-publish
I do think AI has its place - however my overarching opinion is that if you are using AI to analyse a racecard, you obviously want a repeatable process - so tomorrow’s analysis is the same. So if that’s the case, the use case for AI in racing analysis is not the actual analysis - it’s developing the repeatable process programatically.
The big issue with pure AI analysis is that it’s hugely expensive to create a meaningful backtest of analysis - so if you develop some kind of ai analysis process, running that process over the last 2 years of races will cost a serious amount in tokens. You can test it on live races, but how long do you need to build up a meaningful sample to determine if it’s working or not?
https://open.substack.com/pub/hoofs/p/i ... st-publish
I do think AI has its place - however my overarching opinion is that if you are using AI to analyse a racecard, you obviously want a repeatable process - so tomorrow’s analysis is the same. So if that’s the case, the use case for AI in racing analysis is not the actual analysis - it’s developing the repeatable process programatically.
The big issue with pure AI analysis is that it’s hugely expensive to create a meaningful backtest of analysis - so if you develop some kind of ai analysis process, running that process over the last 2 years of races will cost a serious amount in tokens. You can test it on live races, but how long do you need to build up a meaningful sample to determine if it’s working or not?