It would certainly be interesting to see a statistically significant set of results.
Others I’ve seen, so far, have all been loss-making.
The reason, I think, must be that AI is “trained” (as the AI people call it) on information already on the web?
With subjects like forex trading, in percentage terms loads of the information already on the web is actually wrong.
Some of it’s even deliberately wrong, as it’s propagated by various commercial interests.
Most of the rest’s inadvertently wrong just because it’s been copied and repeated - without malice, but just by people who are themselves uninformed and simply know no better.
Since that’s how AI has been trained, it would only be an optimist who’d expect its advice to be of much value?