the shape of things to come

[Big Bad Barney]

I would think developing in-house AI would be prohibitively expensive?

It's not expensive, there's trainable Open Source AI.

If you go into ChatGPT and select Explore GPTs there's hundreds of GPTs that have to been developed by the community.

ahh ic, I guess I orta check that out more thoroughly. Do these models change? What strikes me as problematic is a black box scenario with changing outputs.

Would you know the names of the open source AI's that your typical quant Betfair guys use?

A while back while I had a Pluralsight subscription I skimmed a few of the Microsoft based AI engine building stuff. It has all but gone from my brain, but the only thing I remember is that it was another level beyond mastering regular application engineering. (like anything one learns, there are steps, and there are mistakes one must make to become proficient). It's no small feat. I put it in the do later basket, and it's still there

I suppose the other point is one must know how AI works at a deep level for it to be useful to trading applications. I.e. It's never going to be as simple as going to chatGPT and going, "hey model v51, give me a strategy that makes the big bucks." .... The bar for success there is much higher.

Just now reading how LLM's work (in chatGPT) it seems to me that they are specifically designed NOT to think outside the box.

[ShaunWhite]

Would you know the names of the open source AI's that your typical quant Betfair guys use?

Don't know any that use them, it's not really necessary because it's all regular statistical analysis techniques, maybe some ML. Only thing I use if for is some data processesing but only what I ask it to do not 'find me a strategy' stuff. And generally discussing types of analysis that I'll then do myself. To date it's not great with big datasets, it's struggling at 50mb and I've got 6TB of price data.

Basic ChatGPT is fine without the need for specific training.