as I understand it, this chatbot was trained on data ending in September 2021, which of course severely limits its capabilities. But nevertheless, if you make a copy of it with an updated database, then its capabilities will increase. So why can’t you make it so that every hour/day/month/etc it copies (reproduces) itself with an updated database, thereby constantly improving both in breadth and depth.
Is this a technical or conceptual issue?
If I said something stupid, then I will be glad to anyone who will explain to me why.
at this moment, no one knows what ai can or will do.
if we do not constraint the ability of AI, no one knows what evil that AI will do. So for the sake of human being, do not let AI upgrade itself is a way to save human being in the future. Of course by limited the ability of AI also limited the real power of AI. But lucky we do have a final control, the power button, so we can test AI run without constraint in close environment to test its real ability.
Because training such a large volume of data for LLMs like GPT3 and GPT4 required months of processing and many millions of dollars.
That is the core reason.
If there was infinite computing power and unlimited financial and computing and human resources, the models would be more up-to-date.