Well, I have been trying almost for an year with original GPT when even chatgpt was not available to public and then, came CHATGPT. Also, now I have access to GPT-4 paid version and GPT-4 has been denying continuously that it can't defy financial markets and it's inherent random nature. So all the hype going around these days over internet is different and practical application in forex market is different.
So far it has even failed to develop a reasonably profitable EA over a certain back testing period. Once it can develop a highly profitable EA, then we can consider that we have reached AGI('Artificial General Intelligence'). But yes, GPT-4 is much more advanced than previous versions and it can develop some EAs and sometimes it can make profit as well which can not meet at all to any standard expectations of a trader.
Even the more sophisticated methods like 'MuZero' of DeepMind works quite amazingly well during trained data, but fails terribly during non-trained data of backtesting period. It's a 6 years journey for me and my team and so probably beating forex market completely is the last thing AI can do before reaching AGI.