YATT - Yet Another Turing Test

[smatovic]

[...]

I disagree. Humans have a certain mental capacity and a certain mental performance, there is a limit.

--
Srdja

[chrisw]

[...]

I disagree. Humans have a certain mental capacity and a certain mental performance, there is a limit.

--
Srdja

He’s also got it wrong that the explainer has to do all the work. That’s nonsense, the receiver has to do the work in order to understand. That’s why your teacher tells you bla di bla about history and tells you to go away and write an essay about it. Or the maths teacher shows you some algorithm and then gives you a bunch of examples to work on. It’s doing that work that brings the understanding, just raw passive receiving does nothing.
And doing the work is probably less connected to intelligence than to personality, determination, getting repeat dopamine rewards because repeating success etc.
Hence, btw, passive receipt via the TV or regurgitating ChatGPT does nothing for you, it gives the semblance you know something but in reality is dumbness. As US and western electorates prove again and again.

[towforce]

I disagree. Humans have a certain mental capacity and a certain mental performance, there is a limit.

Your fellow German, Sabine Hossenfelder, who has a high intellect and a large YouTube following, is in agreement with you: in today's video, she points out that a vital part of the brain has a horrendous data throughput bottleneck that even Elon Musk's brain chips won't fix:

[smatovic]

[...]

I disagree. Humans have a certain mental capacity and a certain mental performance, there is a limit.

--
Srdja

He’s also got it wrong that the explainer has to do all the work. That’s nonsense, the receiver has to do the work in order to understand. That’s why your teacher tells you bla di bla about history and tells you to go away and write an essay about it. Or the maths teacher shows you some algorithm and then gives you a bunch of examples to work on. It’s doing that work that brings the understanding, just raw passive receiving does nothing.

Agreed, I had to figure AlphaBeta pruning with pen and paper to grasp it as example.

And doing the work is probably less connected to intelligence than to personality, determination, getting repeat dopamine rewards because repeating success etc.

Agreed, highly skilled underachievers for example.

Hence, btw, passive receipt via the TV or regurgitating ChatGPT does nothing for you, it gives the semblance you know something but in reality is dumbness. As US and western electorates prove again and again.

Agreed, studies show mental degeneration since 2010s with smartphone/internet/social-media, now studies show mental degeneration by ChatBot usage...

--
Srdja

[chrisw]

[...]

I disagree. Humans have a certain mental capacity and a certain mental performance, there is a limit.

--
Srdja

He’s also got it wrong that the explainer has to do all the work. That’s nonsense, the receiver has to do the work in order to understand. That’s why your teacher tells you bla di bla about history and tells you to go away and write an essay about it. Or the maths teacher shows you some algorithm and then gives you a bunch of examples to work on. It’s doing that work that brings the understanding, just raw passive receiving does nothing.

Agreed, I had to figure AlphaBeta pruning with pen and paper to grasp it as example.

And doing the work is probably less connected to intelligence than to personality, determination, getting repeat dopamine rewards because repeating success etc.

Agreed, highly skilled underachievers for example.

Hence, btw, passive receipt via the TV or regurgitating ChatGPT does nothing for you, it gives the semblance you know something but in reality is dumbness. As US and western electorates prove again and again.

Agreed, studies show mental degeneration since 2010s with smartphone/internet/social-media, now studies show mental degeneration by ChatBot usage...

--
Srdja

AB good example, although there are several depths of understanding AB which I would guess GEN Z never bothers with, it's just a line of C (or whatever) that gives Elo. Why? Who cares etc. Underlying understandings get lost, and why not, they're kind of redundant.

Mental degeneration? Argument by Youtube video. They lose the power of speech, but, so? Replace own speech with ChatGPT.

[smatovic]

What happened before...

The first test candidate was posted by Thomas Zipproth, December 08, 2022, generated by GPT-3:

Provide me with a minimal working source code of a chess engine
forum3/viewtopic.php?f=2&t=81097&start=20#p939245

[...]

Second test candidate was posted by Darko Markovic, June 08, 2024, generated by GPT-4o:

GPT-4o made a chess engine
viewtopic.php?t=83882

[...]

Third test candidate was posted by Werewolf, August 14, 2025, generated by GPT-5 Pro:

The Dawn of Skynet
viewtopic.php?t=85356

[...]

[smatovic]

Fourth test candidate was posted by Giovanni Lavorgna, September 23, 2025, generated by Gemini 2.5-pro:

GeminiChess, an LLM built engine
viewtopic.php?t=85490

- C++17 source code
- UCI protocol
- Bitboard board representation
- Magic Bitboard pseudo legal move generator
- AlphaBeta search as negamax
- move ordering with MVV-LVA and SEE
- Quiescence Search with Stand Pat
- check extension
- repetition detection
- Iterative Deepening Framework
- single thread search
- time control
- eval with material, passed pawns, PSQT middle and endgame
- Zobrist hashing
- Transposition Tables
- Killer and Counter Move Heuristic
- Null Windows
- Razoring
- Null Move Pruning
- Internal Iterative Deepening
- Late Move Pruning / Futility Pruning

"Vibe coded" over a weekend, automated via interactive test scripts.

--
Srdja

[towforce]

1A) With hand-crafted eval. 1B) With neural networks.

2A) Outperforms non-programmers. 2B) Outperforms average chess-programmers. 2C) Outperforms top chess-programmers.

3A) An un-self-aware AI, the "RI", restricted intelligence. 2B) A self-aware AI, the "SI", sentient intelligence.

4A) An AI based on expert-systems. 4B) An AI based on neural networks. 4C) A merger of both.

The Turing test really asks for a machine to pretend to be human - which is setting the bar far too low on the intelligence scale.

Your above tests, apart from the sentience test, aren't really Turing tests, but they do make a lot more sense in terms of testing a chatbot.