I don't understand what you're trying to say at all, and probably no one else does either, since no one here is betting on it 🫤

AGI's consensus definition reaches alignment as follows: current tests + new tests in planning+reasoning capabilities

I don't understand this.

@singer the thesis would be broadly as follows:

1) current common evaluations are good for specific skills/knowledge based narrow intelligence assessments

2) seems like biggest perceived gap on LLM-based capabilities are within reasoning and planning, even when scaffolded with helpful cognitive prompting architectures like chain-of-thought or multi-agent systems
3) there doesn't seem to be evaluations in the current eval mix that would look at tests related to either planning or reasoning
4) there seems to be consensus on the ether that AGI has two definition classes (narrow/quantifiable and broad/qualitative), where the broad seems to be referred to the non-scientific perception of "this feels like AGI"

so when 2 would change and 3 would evolve, could then both cases of 4 be commonly agreed upon to be true, more or less both amongst scientists and culturally.

🤷‍♂️🙂🙃🕺

Your explanation does not make the question clearer, sorry.