Increasing model size is only the most obvious way of improving on LLMs. There are many ways of changing LLM architecture and combining them with models from other fields.

I for one am excited to hear of lots of new LLM discoveries and applications, even if they’re not guaranteed.