AI

[imref]

Google just introduced a new AI engine capable of understanding visual images.  This video is simply incredible:

[JCA-CrystalCity]

would it get the message if I gave it the finger?

[English Natsie]

With all of this talk, about AI, I can't help but think that it might be useful if we could find some real intelligence first...  

[imref]

would it get the message if I gave it the finger?

i'm sure it would, and then it would have you eliminated as a threat.

Worth a read: https://cepr.org/voxeu/columns/ai-and-paperclip-problem

[JCA-CrystalCity]

Sounds like Hydra's Dr. Zola's algorithm.

[dracnal]

We've got a LONG way to go before AI is much of a threat:  https://www.aiweirdness.com/optimum-tic-tac-toe/

It's being marketed in a specific way to seem really impressive, but a big part of the issue is it's demoing content that doesn't really require much past cribbing from a couple thousand public examples of something and imitating it. Janelle Sharp, who runs the website linked above, does a lot of fascinating work and talks. Her TED talks are great, and she's a Fellow at the Smithsonian for her work on helping explain the problems with AI using a LLM approach. A specific Neural Net program like AlphaGo that is only focused on one problem can get extremely good at it. Something that's focused on trying to be broad and generalized, is not. It's hilariously 'ask a toddler  to explain science' bad.

[imref]

OpenAi has released GPT 4-o, which now can understanding and converse in voice and video, as well as chat. The demos are amazing and unlike the Google demo I posted to kick off this thread, it wasn't staged.