In retrospect, the flaw in the Turing Test was always using human discernment as the model to measure against.
LLMs easily pass the Turing Test, not because they’re sentient, but because we’re dumbasses who interpret the most dopamine-supplying response to ourselves as the most intelligently human.
@Catvalente I think the Turing Test was intended as a necessary condition required to demonstrate artificial intelligence, but not necessarily a sufficient condition.
To be fair, there are probably people who cannot pass a Turing Test.
@wollman @StryderNotavi @Jvmguy @Catvalente
Exactly. We’re mistaking pattern mimicry for reasoning — like seeing faces in clouds and calling them sentient.
The danger isn’t that these systems “decide” anything. It’s that we hand them power because they sound like they do.
Surveillance. Targeting. Soon, weapons.
Not intelligence — just simulated judgment wrapped in UI.
Anthropomorphism is a bug — a potentially deadly one.
@GeePawHill @Jvmguy @Catvalente @jasongorman I think you might be right.
@StryderNotavi Maybe you saw it here: “People see faces in the side of a mountain or a piece of toast, and in the same way perceive the output of LLMs, mistaking cogent-sounding statistical probability with intelligence.”
@davel Looks like a few people have made the observation, but @GeePawHill rings true here - https://mastodon.cloud/@jasongorman/114595098303670564
That's the post I remember.
@StryderNotavi @Jvmguy @Catvalente
I book-marked it for you :)
@Jvmguy @Catvalente well, Turing test was created in the period in psychology that behaviorism dominated it, with an iron fist.
The idea was that any talk of emotion (any at all) was subjective. Babies don’t get upset and cry. They make loud noises in periods of not eating.
Also consciousness, itself, is really not measurable.
So for someone that wanted to operationalize the whole thing at the time “if it talks like a duck and walks like a duck… “ was probably the only option.
The Turing test was not intended as a test, but as a rhetorical ploy.
It was a sidenote by Turing to illustrate the concept to his peers.
Thus, it's no surprise, it does not measure what people think it measures.
@jon @Catvalente It’s being put into use whether we want it or not. Given that, I guess I’ll hope for the best, but I will not be surprised at all if it leads to the collapse of civilization entirely.
An exciting time to be alive.
I will always maintain that the Turing test is not a serious piece of computer science, but a product of Turing living in a society with a lot of expected thoughts and behaviors. The argument runs "if some subject can respond to a human's questions such that that the human cannot tell that they are different, then it is not truly different." That makes no sense about a computer, but makes perfect sense for someone deep in the closet to try and find comfort in it.
Anytime!
Honestly, I can't see the Turing test as more than a metaphor to get people over the hump of rejecting the very idea of machine rationality.
I mean, one of his "pass" criteria was the use of psychic powers by a computer. This is not science, it's marketing.
If that’s true that’s like 249 possible biases not including the implicit learned ones. So calculating that diversity that we don’t even realize exists. Sounds like an adventure in experience and self I guess.
@Catvalente Also, the Turing test was never about a machine *being* intelligent - it's about a machine *convincing a human* that it is intelligent.
Which, as we now know far better than Alan Turing could have ever imagined, is tragically easy to do.
@Catvalente people have given way too much credence to the Turing test, it is a case of faulty logic.
If intelligent beings talk a certain way, then talking a certain way shows you are an intelligent being?
Genius artists can replicate a view realistically. Cameras can replicate a view realistically so cameras are genius artists?
Intelligent people are good at chess, AlphaZero is good at chess, so AlphaZero is an intelligent person?
Causation doesn't work like that
I always thonk that Turing test was hypothetical like Schrodinger cat or Russel's teapot or trolley problem
I thonk so, too.
@Catvalente @Gargron
Hmm 🤔
If I cross the streams, the Turing test should be done in part by the most intelligent bears 🐻.
While they are away the Pope can use the woods.
🤷🏻♂️
@Catvalente it was never intended as a test of sentience or intelligence, but a challenge to our reflexive dismissal of the ability for computers to think.
From wikipedia:
LLMs do think and possess limited intelligence (but not human intelligence, think more like cockroach intelligence).
@Catvalente terry Winograd wrote about this (as did hofstadter) two decades ago. You are absolutely right.
See https://berryvilleiml.com/bibliography/ for pointers.
@Catvalente It reminds me of this:
Perhaps our mistake is in thinking that we are sentient.
@Catvalente
Turing test was always a thought experiment, but your solid point does make me think- have we actually tried the mental acuity tests?
I genuinely don't know if they would work, but the tests where they ask a number of things about current events, remember these 4 words, could you talk about this subject, repeat these things back to me...
I actually am not sure.
Catherynne M. Valente
Kevin Morgan
Stryder Notavi
Garrett Wollman
qurlyjoe
AnthonyCFox
Everyday.Human Derek
GeePawHill
Jason Gorman
davel 🚫
filobus
Death by Lambda
Perma
Ville Takanen
jon
Raglan Niall 
James M.
Dave Rahardja 26
Marty
PhDog
Danny Lucas
Cassidy Curtis
Resuna
Carl Muckenhoupt
nigel
Grovewest
xander
JWcph, Radicalized By Decency
Tom Lowe
Askold
Walrus 🏴
The Animal and the Machine
Shiri Bailem
MegatronicThronBanks
noplasticshower
Charles Fulton
Michael Westergaard
Orjan
VZ
Jerry Argyriou
Personne intéressante
Hardhead13
Oggie