I think it is AGI, seriously. Try asking it lots of questions, and then ask yourself: just how much smarter was I expecting AGI to be?
As I’ve argued in the past, AGI, however you define it, is not much of a social event per se. It still will take us a long time to use it properly. I do not expect securities prices to move significantly (that AI is progressing rapidly already is priced in, and I doubt if the market cares about “April 16th” per se).
A place of politics, culture, and discourse
The benchmark that makes sense to me: Can it beat Pokemon?
10-year-olds can beat Pokemon.Report
An exchange I just had with ChatGPT:
Me: Do you know what time it is?
ChatGPT: yeah, it’s time for you to get a watch 😎
AGI confirmed.Report
Artificial General Snarkiness. But what model were you using?Report
The new one. So far, it’s pretty good.
I still agree with the Pokemon test for AGI but this thing is getting to be seriously uncanny.
The argument that “intelligence must be X to be AGI” strikes me as an argument that eliminates more humans than is comfortable to exclude ChatGPT.
But if you want a bright line, it remains Pokemon.Report
Yeah it’s not well defined enough for me to even worry about – I just continue to be amazed and a little worried at the progress. Right now is a great stage where it’s making our lives at work easier but hasnt gotten to the point where the humans are being fired – though there’s definitely a lot of pushback on new hires or even backfills while we’re learning just how much productivity we’re getting from these tools.Report
Back in 2023, I failed the Turing Test for a few minutes before figuring out that it was just auto-complete with some extra steps.
And then you do stuff like contemplate consciousness and try to define it narrowly and then fail and then try to define it broadly and then fail and do some research and the thought “am *I* just auto-complete with a lot of extra steps?”
Anyway. The attempts to define any given concept wrt AI will eliminate humans. It’s fine to eliminate, oh, up to 5%, I’d say. Hey, some folks just roll natural 1s.
Once we start defining stuff so that we’re talking about 10% of people (or higher), I start to get nervous.Report
As for AGI, what does the software do when no one is bothering it with questions or tasks? Does it consider the things it might spend cycles on, and choose from among them? As it does things, does it consider what it might add to, or subtract from, the set of things it might spend cycles on? Does the set include not answering the questions or assigned tasks? If not, it may be a clever tool, but it’s not AGI.Report
Not to nitpick, but that excludes more than 5% of humans.Report
Anyone who can’t pass that test shouldn’t be let out without a keeper.Report