i believe it won't be possible very soon, but within the next decade? yeah, considering how much progress is being made on current models

So if there's a transformer anywhere in the network, this resolves no? Do you mean LLM rather than transformer? What about an LLM with an RL agent running outside it to provide agency, or vice versa?

@Tomoffer If the core functionality of the AI stems from a transformer, this resolves no. I realize this is a bit subjective but "I'll know it when I see it". Essentially, this only resolves yes if there's a breakthrough model that beats current transformer-based models in a significant number of benchmarks.