Gunjan Banerji

Professor Lowe said this to me, and I found this as well, which is surprisingly, a lot of these AI models are not that good at arithmetic.

And I actually found that ChatGPT made a math error very early on in my process that led me to kind of misallocate funds in this made-up scenario.

I used a stock market simulator to track some of the performance of these trades and moves that ChatGPT was suggesting, and I did find...

that it initially kind of misallocated how much money I would be leaving in cash.

So one, definitely double check the math.

You know, small numbers can have really big implications on investment returns.

So make sure you're doing your due diligence there.

I would also advise people to put these models or at least ChatGPT into thinking mode rather than auto.

And then during my experiment, I started putting ChatGPT into candid mode and that can lead it to be slightly less sycophantic.

this tendency for some of the models to tell you what you might want to hear.

That's markets reporter Gunjan Banerjee.

Gunjan, thanks for your time.

Thank you.

This has been great.