Conversation

Notices

1. Embed this notice
   Rob Landley (landley@mstdn.jp)'s status on Tuesday, 27-Aug-2024 00:18:02 JST Rob Landley

   The Turing Test is like the Bechdel Test: quick smoketest to eliminate things that aren't worth a closer look.

   The fact either filtered out most entrants for so long is a damning indictment of the state of the industry.

   • clacke likes this.
   • Embed this notice
     Rob Landley (landley@mstdn.jp)'s status on Tuesday, 27-Aug-2024 00:18:03 JST Rob Landley

     in reply to

     • Charlie Stross

     @cstross Agreed. I've never seen it well defined, it's always just "does it fool you personally for 15 minutes". Eliza did that to people in the 1970s. The Sims did it here:

     https://aliceandkev.wordpress.com/

     People assume there are professional Turing testers, a man with a clipboard somewhere who's an expert at testing turings and you can phone them up to make an appointment. Meanwhile, humans see faces in clouds, put googly eyes on the roomba, ask their car out loud to start on cold mornings...

     clacke likes this.
   • Embed this notice
     Charlie Stross (cstross@wandering.shop)'s status on Tuesday, 27-Aug-2024 00:18:04 JST Charlie Stross

     in reply to

     @landley You just hammered my knee-jerk reflex: THE TURING TEST AS WIDELY UNDERSTOOD DOES NOT EXIST. (Ahem.)

     Turing's original paper "Computing Machinery and Intelligence" set up the thought experiment really badly: it's really a test of whether a chatbot can emulate performative feminity—gender-coded language. It says more about Turing (gay Englishman from a quasi-monastic single-sex background) than about AI.

     (You're not wrong about the "first cut" thing, but …)

     https://courses.cs.umbc.edu/471/papers/turing.pdf

   • Embed this notice
     Quinn Norton (quinn@social.circl.lu)'s status on Tuesday, 27-Aug-2024 03:23:13 JST Quinn Norton

     in reply to

     • Charlie Stross

     @cstross @landley it feels like this thing where we hold people to a too high standard. Turing is allowed to have stupid moments like the rest of us without people spending decades of efforts trying to figure out how that one idea was brilliant even though it obviously was really dumb.

     Alexandre Oliva likes this.
   • Embed this notice
     nonlinear (nonlinear@social.praxis.nyc)'s status on Wednesday, 28-Aug-2024 16:07:52 JST nonlinear

     in reply to

     • Charlie Stross
     • Quinn Norton

     @quinn @landley @cstross the issue is that when there’s a metric, it becomes the goal.

     clacke likes this.
   • Embed this notice
     Quinn Norton (quinn@social.circl.lu)'s status on Wednesday, 28-Aug-2024 16:07:53 JST Quinn Norton

     in reply to

     • Charlie Stross

     @landley @cstross I am aware that Turing was very smart, and that he was gay, and chose chemicals over prison. Thanks. I am, in fact, blaming Turing for having a dumb idea in the Turing test, actually a very dumb idea. I think it's ok that he had a very dumb idea, I just wish people could accept that it was very dumb and stop bringing it up.

   • Embed this notice
     Rob Landley (landley@mstdn.jp)'s status on Wednesday, 28-Aug-2024 16:07:54 JST Rob Landley

     in reply to

     • Charlie Stross
     • Quinn Norton

     @quinn @cstross Turing was breaking new ground in 1950. The first computer to use transistors started construction 5 years later, and Fortran shipped 2 years after that.

     Before either of those the British government chemically castrated Alan Turing for being gay, who then committed suicide in 1954.

     Nobody's blaming Turing.

   • Embed this notice
     Rob Landley (landley@mstdn.jp)'s status on Wednesday, 28-Aug-2024 16:07:58 JST Rob Landley

     in reply to

     • Charlie Stross
     • Quinn Norton

     @quinn @cstross "necessary but not sufficient" is not the same as "dumb".

     Golems were pottery, then irrigation happened and now intelligence was hydraulic engineering with "humors". When we moved to clockwork it was the mechanical Turk and TikTok of Oz. Then electricity showed up and powered Frankenstein and Metropolis (1927). Then we got Colossus and the Harvard Mark 1 actually implementing a difference engine and Turing asked what a scientific approach to the age old question might look like.

     clacke likes this.