Center for Artificial Artificial Intelligence

A new initiative designed to bring together all sides in harmony.

AI can’t really think because it is artificial.
Just backprop or simple rules.
And you are so irresponsible for suggesting that a machine could do lots of different things quite intelligently.
And Theranos, and blockchain.
And something really, really vague about quantum computing from Penrose
And some examples of things ChatGPT got wrong last year. Oh wait, most are fixed already.
And some upset people from vague fields whose definitions never really helped anyone build anything useful.

Humanity is as dumb as a rock so you have an obligation to avoid them getting confused…
… by all the people claiming that more than “sparks of AGI”.
… except they aren’t. Nobody can actually point to a single paper making outrageous claims, more than “sparks of AGI”.
But its a massive hype problem, clearly …
… caused by “computer scientists, statisticians, or other abstraction-and oversimplification-oriented individuals” (source).
For example the 100 million users of ChatGPT are presumably all idiots.

Things that act intelligently across a stunningly broad spectrum of tasks can be labelled AAI without the risk of enraging anyone.
Human egos are absolutely safe in all other respects.
Mostly useless fields won’t be forced into contorsion as performance steadily improves.

But it already says “artificial” in AI

Oh right, yes so it does.

It isn’t clear that humans can truly learn. But if you would like to build algorithms that fake it:

Purpose	Time	Procedure
Individual help getting started	11:30am Friday EST	Send a cal invite to peter dot cotton at microprediction.com
General stochastic banter, see below.	Noon Friday EST	See the slack channel for invite and cancellations.

If you can measure it, consider it predicted