Eh it’s the illusion of speed. Scaling brought enormous returns from GPT-3 -> GPT-4 but it’s been far less significant for every major release since. To compensate for this, every research lab is coming up with new ways to extract value of it of models: CoT, RL, Agent Harness etc

However, these are all hacks to make LLMs more efficient or (try) to make them more reliable. They still have significant drawbacks which will take years (probably decades) to ever get them to the point where they can reliably replace knowledge workers. China knows this and is taking a far different approach to LLM development (not a tankie fyi). Scaling is a horrible idea which will burn billions of dollars with an astronomically low chance of return.

I can agree with that largely but I still contend you’re conflating a few things to make that argument. Fundamentally an LLM will make predictions based on probability (ignoring temperature) and probability does not equal certainty.

It’s temperature primarily. That being said there is still a chance that an LLM can output values that are unexpected even at low temperatures.

What are you saying precisely? It’s well known that LLMs have non-deterministic output (Ilya Sutskever even claims as such). Are you saying the way it goes about retrieving tokens as deterministic?