Jamey Sharp is a user on toot.cat. You can follow them or interact with them if you have an account anywhere in the fediverse. If you don't, you can sign up here.

I want a series of Fred Rogers for President campaign ads. The Cabinet is sitting in the President's briefing room, outlining today's crises and concluding with a plea: "Mr. President, what do we do?" Everyone turns to the head of the table where an old TV and VCR turn on, and as the CRT warms up the VCR fast-forwards to a clip from "Mr. Rogers' Neighborhood" that covers exactly this situation. In one corner a general starts quietly weeping tears of joy, realizing that sharing is an option…

@jamey This is what machine learning should be used for!

@salmon_cannon I was partly inspired by Penn and Teller's "Lab Scam" (youtube.com/watch?v=fxMKuv0A6z) which demonstrated the aesthetic I'm imagining. The buzzword they built that skit around was "voice recognition" rather than "machine learning", but it was very much the same idea. 😁

@jamey this was terrible. I apologize, but if we could teach a machine to love based on Mr. Rogers, I think we will win

@jamey But does that mean that we'd first need the computer to be self-aware and insecure to need the reassurance of Mr. Rogers? πŸ€”

@salmon_cannon Hah, I don't think intelligences have to be insecure to get value from learning to love. ❀️

@jamey I really wish I had the programming chops to make a ML algorithm with Mr. Rogers eps as source.

@salmon_cannon An Eliza-like bot that responds with Mr. Rogers clips would be amazing too 😁

I feel like you could get something that would fool people into feeling like it was deeply insightful by just hand-selecting a bunch of clips and picking key words that trigger each one πŸ€”

@jamey I'm almost thinking of something that basically remixes Mr. Rogers' quotes into unique things that apply to the given situation. But that might be way beyond current computing power.

Jamey Sharp @jamey

@salmon_cannon I dunno whether it's being current computing power, but it's definitely beyond my capabilities!

I think people have actually demonstrated all the pieces you'd need to put together for at least a half-assed version of that (simulated faces and voices trained on past video, plus natural language interfaces). I also think that would be super creepy to use this way. πŸ˜…

Β· Tusky Β· 0 Β· 1

@jamey
Oh yes it would be for sure deep in the uncanny valley