This is what PI AI says to limit liability:

" Pi may make mistakes, please don't rely on its information." (this is placed below whatever answer PI AI gives at the very bottom in fine print).

So, the problem with most human beings is that they expect AIs to be factual. They are NOT factual. They are only prediction engines. 

If you look at AI like you do weather reports you will be fine. It's sort of like they are predicting the weather.

Devices like AI for example, could help People who invest in Stocks make better choices but only MAYBE.

Devices like AI might help people who are professional gamblers of one kind or another (OR NOT).

IT"S sort of like a weather report.

How many times have weather reports been wrong?

I might say at least around 50% of the time.

The same might be true  of AIs too.