It will probably be a bit of both. GPT-4 anecdotally hallucinates singificantly less than GPT-3.5. This can be seen because often hallucinations that exist with GPT-3.5 do not exist with GPT-4 whereas I have never experienced the reverse.

If the trend continues at this rate, then hallucinations should get less common but it will be decades before they can really be trusted.