How close are we to AGI and do people actually understand what that means? Too often people use the term as if General were 'ubiquitious' as in a broad application of AI instead of one that is inditinguishable from human intelligence.
For me, "general" is a sliding scale, not a yes-or-no boolean, so I think it's fine to say that GPT3, let alone later versions, is "general" in the domain of text — it has more general knowledge than I have, that's for sure!
Likewise "intelligent", though for me that's a vector value: [i_0, i_1, …, i_m] where each i is the ability at some aspect of intelligence.
In fact, I'd go further and say that you can define "generality" with respect to such a vector: if the intelligence vector is a normalised score 0-1 for each value, then generality could be easily defined by the statistics of that array.
As even GPT-4 seems to be better than the layperson and worse than an expert at everything (textual) it has been tested on, I'm not sure if it's more or less general than a human by this standard.
And you may wish to define intelligence by ability to learn from small data sets rather than large, at which point we then need to consider the difference between the initial training, the RLHF, and any fine-tuning; and how this compares to evolution, education, and work experience.
no because it means something different to each person (both its definition and its implications)