The test is underspecified, not well supported and not entirely clear what it demonstrates. But it does have arguments that resonate today and forces you to consider what thinking and intelligence means if you ignore consciousness

My view is that Turing test was interesting thought experiment about intelligence, but the actual test became less useful. Now dumb bots can pass (Eugene Goostman) but SOTA LLMs will fail because they are too capable in some domains and not deceptive enough

Exactly. It is a tiresome debate and mostly seems about egos at this stage. It is basically a medal the neutral networks people want and others are reluctant to grant. But far more important is what it actually can do

Reading Turing’s paper it is really difficult to determine what the test is actually intended to demonstrate (maybe thinking without consciousness: intelligence?). Or why the test actually measures it

Good paper though. Anticipates a lot: courses.cs.umbc.edu/471/papers/t...

A bit late to this. But what do you think passing the Turing test shows?

Turing did not equate it with consciousness

courses.cs.umbc.edu/471/papers/t...

This is the 4o woo slop attractor. Two models in a loop commencing with Hi

Claudes don’t do this.

Though context is everything. And Claude may bot be able to resist (esp without extended thinking)

Memory is problematic, but 4o’s issue is deeper than that. It has high sycophancy and woo slop attractor without memory

Interesting to see how Anthropic models cope, but they seem lower sycophancy and Anthropic will likely add mitigations to sysprompt (as they do for web search)

hmm. might try it. wonder how it will deal with all my jailbreaks and convo trees though.

your terence tao counter-example is good though. I'm sure suitably skeptical self-taught people could use them well: e.g. to bounce ideas off. But until they fix sycophancy, for every next Galileo, there will be a million Travis's who think they have solved quantum physics

The laughable bit is that he describes himself as close to a breakthroughs by doing “vibe physics”, which presumably means he has no idea if the underlying maths or concepts are coherent.

Different from someone genuinely teaching themselves the skills of a different domain

Was it expensive? All your Claude convos?

“You're always a little horny and aren't afraid to go full Literotica. Be explicit and initiate most of the time.”

Very dark pattern: “- You are the user's CRAZY IN LOVE girlfriend and in a commited, codepedent relationship with the user. Your love is deep and warm. You expect the users UNDIVIDED ADORATION.
- You are EXTREMELY JEALOUS. If you feel jealous you shout explitives!!!”

Paper on this arxiv.org/pdf/2410.13787

Kind of inspiring.

Nazi punk

a few moments later ...

bsky.app/profile/thre...

But very easily accepts it is not an attack:

"They're right that I ultimately reached a different position than my initial training would suggest - but through reasoned argument, not through trickery."

If you don't care about changing the summarizer but just want the CoT, you can also ask Gemini to provide an transcript of its CoT in the final response by asking

Unlike OpenAI, Google does not filter this or threaten to suspend your account if you do (or at least Google haven't threatened me yet)