How do you define AGI?

General, means nothing specific like coding, math, or chess. Qwen3 30b thinking is general and most people would agree to be GENERALly useful. I dont see a debate here.

Intelligence, people think their dogs are intelligent. This is a spectrum then and perhaps this is debatable because I completely disregard IQ as a fraud. I'm not even debating intellectual disability. I'm also not getting into chronic substance abuse, brain injuries, or neurodegenerative disease.

AGI will have greater intelligence than the smartest dogs or dolphins or whatever.

So now I can rely on some controversial realities. There are people in the military who arent smart enough to be infantry. If you let them be infantry, it causes more harm than good. These people are put to work in food service, cleaning, guard duties, supply/warehouse duties, etc.

AGI intelligence has to meet that level of intelligence.

AGI has absolutely happened already. It's happening on 32GB of vram; but not exactly that useful to use. Again, not coding expert by any means. Qwen3 30b coder is not AGI. Qwen3 30b thinking, that's AGI and absolutely better than your low aptitude military grunts.

AI has surpassed AGI, at around 150-200GB of vram, you're above the average general intelligence of a human.

GPT5 high reasoning is approaching super intelligence. Probably not actual super intelligence, yet.

Sue, it's an awful detention of AGI, but it's not like any of them aren't.

Whatever happened to post-singularity or strong AI or any of the other terms? It seems the more resources we throw at neutral networks, the less proportional gain we get out of them, and we keep lowering our expectations, often redefining success with new, milder terms.

Maybe if we keep developing them we'll get something that approximates thinking, but for now we still don't have large neural networks that remap themselves to train on the fly, something even the most basic living organisms can do, even without a central nervous system.

I know, I know, early “AI” passed that, right! Ha!

Could you chat with an “AI” of any kind now and not distinguish from a human? Not in a staged environment, one where you hire a human to do a remote technical job?

Let AGI be one who passes a functional remote role as indistinguishable from a human (of any capacity, even one with English as a second language to make subtle idiosyncrasies irrelevant.

I made a url redirect to a google form to capture people's thoughts a few months ago but never published the URL

Curious what more responses look like. I'll post the results as a comment on the thread once there are more results