If we wonder how Large Language Model (LLM) providers will try to improve their services in the next few years, we can start by trying to foresee how current limitations will be addressed. While LLMs have been fairly successful in the chat box format, they are both expensive in terms of energy usage and have interminable problems with hallucinations. Software developers battle with increasing token usage to achieve more focused results.
There is still a bit of guess work involved in working out how size vs. training truly effects output, but…
