I'm a bit sceptical of the "no moat" proposition because (a) ChatGPT 4.0 really does seem in a different league and (b) it's clearly very hard to run. I haven't seen anything from the explosion of open source / community efforts that comes close for general applications.
The take in the post rings of the classic trademark Google arrogance where they assume that if somebody else can do it they can do it better if they just try - where the challenge of "just trying" is discounted to zero. In reality, "Just trying" is massively important and sometimes all that is important. The gap between unrefined model output and the level of polish and refinement that is apparent with ChatGPT 4 may appear technically small but it's the whole difference between a widely applicable and usable product and something that can't be more than a toy. I'm not sure Google has it in it any more to really fight for something they want to achieve that level of polish.
Just wait a few months. You are underestimating thousands of researches and engineers only working on this with enormous compute budgets in several companies.
Version 4 also now supports 32k tokens, good luck handling that on even awesome gaming local dev rig machines, although perhaps with linformer ideas, block-wise algorithms to handle larger than GPU memory, universal memory / RDMA it’s entirely doable. I got 50,000 atoms simulation back in 2018 on 11gb vram, at 32bit floats, the software stack has come a long way and now we have the 24gb 4090 with bfloat16, and vector DBs, and the infinite-context transformer paper just came out, so models all ought to be retrained on that if the method is truly superior anyway, not sure how atoms translate to pages of text but it’s almost surely possible to make a pretty useful LLM.
The take in the post rings of the classic trademark Google arrogance where they assume that if somebody else can do it they can do it better if they just try - where the challenge of "just trying" is discounted to zero. In reality, "Just trying" is massively important and sometimes all that is important. The gap between unrefined model output and the level of polish and refinement that is apparent with ChatGPT 4 may appear technically small but it's the whole difference between a widely applicable and usable product and something that can't be more than a toy. I'm not sure Google has it in it any more to really fight for something they want to achieve that level of polish.