"Voices from DARPA" Podcast, Episode 83: When Should Machines Decide?

[@guykonstrucks8733]

The alignment of how an AI makes decisions being variable, categorizable, and tunable is really wise and an important area of study. Definitely a proper way to go with AI development, since the biggest hurdle in AI building is the inability for us to understand the complexity and larger reasonability's of the neurons in a neural network. That's why we just have to throw in massive amounts of data that's 'close enough' to our target product and use math to refine every bit of it millions of times over. This is like a way to re-wrangle things back into a language we understand. This area of study, making a 'moral compass' of - and potentially for - AIs, is a really important and valuable area of exploration right now.

[@watcher1326]

Such an interesting problem! Loved hearing about these projects. I hope future episodes follow up on the progress.

[@skytunes86]

John Conner's over here all; 🤨 As you're talking about handing more control to AI. . .

[@Wild-t4e]

Idk if you caught my conversation to the Aussies about the 2 hooligans, especially 1 was pretty tricky mean! I could crush them if they just knew they’re nothing compared to DARPA… And then I realized a piece of my adapter is missing! Probably the trashy hooligan took it while still at ESA in Poland…

[@tranquilvortex]

Never. "DAVE: Open the pod bay doors, Hal. HAL: I’m sorry, Dave. I’m afraid I can’t do that." Machines have no spirit or soul. They should never decide except about how to destroy themselves.

[@AngeloPerfili]

That's some weapons grade , project mockingbird style propaganda you got there designed to make all the simps, whimp's and blimps bobble their heads up and down...........................................eh, KZbin?

[@Wild-t4e]

DARPA has a real human AI hybrid in Switzerland! She’s hybrid she doesn’t age…

[@web3Aicache]

Justin

[@AquisQuerquennis]

Bottom line: If you trust a methodically trained soldier made of flesh and blood, trusting a methodically fine-tuned synthetic intelligence shouldn't make any difference.

[@brandondavid3643]

never….ever