Another thread from last week’s BlueSky:
In the AI/anti-AI discourse, I haven’t seen enough treatment of the problem of ‘feeling’. A silicon-based ‘intelligence’ can’t feel as a human does: it can simulate touch, but not experience touch; it can calculate a response to visual stimuli, but it can’t see. And above all, it can’t have what in humans we call ‘feelings’. At best, it can call up a response that it has learned would be most probable if humans were to encounter particular circumstances — a break-up, death in the family, a favourable job offer — but there’s nothing internal (as it were) to the electronics that produces the thing a human feels. Nor, all the more, mixed feelings.
Now, I don’t put any stock in a feeling-mysticism that treats art objects as if imbued with wooooo ‘authentic feelings’ — but it seems vital to me that we distinguish the algorithmic processes that display a probable response similar to what a vast, fine-grained database implies a human would feel, from the human feelings themselves. You can’t surprise a computer. You can’t charm a computer. You can only submit input that you calculate will induce AI to guess at a human-seeming response.
To which Ryan Turnbull responded: ‘If an AI could speak, we couldn’t understand it.
– Wittgenstein, probably’.
Also, my most-frequently-played musicians for May:
1 Echo & the Bunnymen 7 scrobbles
2 Iron & Wine 7
3 The Beatles 7
4 The Who 7
5 Bruce Springsteen 6
6 Elvis Costello 6
7 Nick Cave & the Bad Seeds 6
8 The Rolling Stones 6
9 Van Morrison 6
10 The Clash 5
As always, I omit the Mountain Goats because they’d end up at or near the top every month. That’s partly because I am a big fan, but also because of the vast catalogue they’ve produced and the number of concert recordings I’ve retrieved.