Buongiorno, Daniela Tafani <daniela.taf...@unipi.it> writes:
[...]
> Un'analisi critica della seconda ipotesi di Bengio
>
> ("A computer with human-level learning abilities would generally surpass
> human intelligence because of additional technological advantages")
>
> si si trova qui:
>
> https://nitter.snopyta.org/MelMitchell1/status/1661476453068976130
grazie Daniela per la segnalazione e siccome non vorrei che la sostanza
dell'analisi critica vada persa nel null-verso, la riporto qui verbatim,
/ricucita/ da quel tristissimo modo di esporre le idee che sono i
cinguetti "a singhiozzo":
--8<---------------cut here---------------start------------->8---
I'm reading Yoshua Bengio's new blog post, "How Rogue AIs may
Arise". yoshuabengio.org/2023/05/22/… Mostly it's the same arguments as
in earlier writings by Bostrom, Russell, etc. Lots to say about all this
but there's one issue I want to point out. 🧵 (1/8)
There's something that always strikes me as strange in these
arguments. Bengio frames it as a kind of syllogism: (2/8)
Hypothesis 1: Human-level intelligence is possible because brains are
biological machines.
Hypothesis 2: A computer with human-level learning abilities would
generally surpass human intelligence because of additional technological
advantages.
Claim 1: Under hypotheses 1 and 2, an autonomous goal-directed
superintelligent AI could be built.
The "technological" advantages in Hypothesis 2 are things like speed of
computers vs. brains, ability of programs to clone themselves and thus
"benefit from and aggregate the acquired experience of all its clones",
etc. (3/8)
However, I find hypothesis 2 suspect. Indeed brains are biological
machines, but a very specialized kind of machine, adapted to a specific
evolutionary niche, with many cognitive features (biases) that are
advantageous in that niche. (4/8)
Hypothesis 2 (rather glibly) implies that "human-level learning
abilities" could attained in machines, conveniently separated from all
those human "limitations", and so "obviously" machines with these
abilities would be superior to humans. (5/8)
Doesn't seem at all obvious to me! Our learning abilities and
intelligence are all wrapped up together with our embedding in a social
/ cultural / evolutionary niche. (6/8)
Not clear at all that these abilities could be so easily "sifted off",
so that machines could, in effect, have their cake and eat it -- that
is, have human-level intelligence + all the advantages of
computers. (7/8)
I've written about all this before, but the argument keeps coming up
again and again, and I find it a bit frustrating that Hypothesis 2 is
always taken as a given, with no argument. That's it for now. (8/8)
--8<---------------cut here---------------end--------------->8---
[...]
saluti, 380°
--
380° (Giovanni Biscuolo public alter ego)
«Noi, incompetenti come siamo,
non abbiamo alcun titolo per suggerire alcunché»
Disinformation flourishes because many people care deeply about injustice
but very few check the facts. Ask me about <https://stallmansupport.org>.
signature.asc
Description: PGP signature
_______________________________________________ nexa mailing list nexa@server-nexa.polito.it https://server-nexa.polito.it/cgi-bin/mailman/listinfo/nexa
