Community Archive

🧵 View Thread

🧵 Thread (23 tweets)

Placeholder
Emmett Shear@eshear• 8 months ago

All IQ tests are actually mind reading tests. There are an infinite number of patterns that start 2, 4, 6, ... and the fact humans (almost) all privilege one of them is a fact about humans, not a fact about math.

3.0K 128
4/3/2025
Placeholder
ꜱᴘᴀᴄᴇ ᴘᴜɴᴋ@_space_punk_• 8 months ago
Replying to @eshear

@eshear I mean the one when I was 13 also asked me to guess how turpentine was made and the one when I was 32 also asked me to say as many words that started with the letter A within a time limit but I see what you're saying

37 0
4/3/2025
Placeholder
Emmett Shear@eshear• 8 months ago
Replying to @_space_punk_

@_space_punk_ Yeah I mean you can poke at IQ in a lot of different ways, as you say. But they all come back to mind reading in the end. Even more obvious when it's things like words invented by humans!

59 0
4/3/2025
Placeholder
GoldMagikarp@GoldMagikarp42• 8 months ago
Replying to @eshear

@eshear A test of whether your brain has encoded the blessed inductive priors.

282 0
4/3/2025
Placeholder
Emmett Shear@eshear• 8 months ago
Replying to @GoldMagikarp42

@GoldMagikarp42 Our blessed inductive priors, their vile arbitrary biases.

434 18
4/3/2025
Placeholder
David Zechowy 🖇️@davidzech27• 8 months ago
Replying to @eshear

@eshear continue the sequence *such that total kolmogorov complexity is minimized*

32 0
4/3/2025
Placeholder
Emmett Shear@eshear• 8 months ago
Replying to @davidzech27

@davidzech27 Kolmogorov complexity is non-computable, it reduces to the halting problem. Why is this everyone's answer???

36 0
4/3/2025
Placeholder
Ben Somers@ben_m_somers• 8 months ago
Replying to @eshear

@eshear Name another pattern that starts 2,4,6…

3 2
4/3/2025
Placeholder
Emmett Shear@eshear• 8 months ago
Replying to @ben_m_somers

@ben_m_somers 2,4,6,2,4,6,2,4,6,...

346 0
4/3/2025
Placeholder
Weltgeist@WeltgeistYT• 8 months ago
Replying to @eshear

@eshear @ben_m_somers This is wrong because you're not looking for a pattern within the first 3 digits, which is the assignment According to this logic, there's infinite answers, such as 2-4-6-0-2-4-6-0-2-4-6. You need to establish how 4 relates to 2 as 6 relates to 4

33 1
4/3/2025
Placeholder
Emmett Shear@eshear• 8 months ago
Replying to @WeltgeistYT

@WeltgeistYT @ben_m_somers Correct!! There are infinite answers. Why prefer one over another? (Hint: context)

89 0
4/3/2025
Placeholder
Ben Somers@ben_m_somers• 8 months ago
Replying to @eshear

@eshear @WeltgeistYT I think we "prefer" the pattern that is simplest to explain

3 0
4/3/2025
Placeholder
◌ moon ●@leonine_moon• 8 months ago
Replying to @ben_m_somers

@ben_m_somers @eshear @WeltgeistYT "these numbers over and over again" and "add 2 each time" are IMO about equally simple

1 0
4/4/2025
Placeholder
bashu, thanks@bashu_thanks• 8 months ago
Replying to @eshear

@eshear orgs ship their org chart humans ship the patterns they recognise

22 0
4/3/2025
Placeholder
Emmett Shear@eshear• 8 months ago
Replying to @bashu_thanks

@bashu_thanks Patterns ship their patternself

24 1
4/3/2025
Placeholder
Matthew Pirkowski@MattPirkowski• 8 months ago
Replying to @eshear

Or it’s just the straightforward Bayesian inference. How many times has “detecting evenness” proved useful compared to the rest of those patterns. The Bayesian perspective—as opposed to Frequentism—bridges elegantly between mathematical formalism and the non-ergodic reality of organismic intelligence that must navigate a coarse-grained and multi-scale adaptive landscape.

52 0
4/3/2025
Placeholder
Emmett Shear@eshear• 8 months ago
Replying to @MattPirkowski

@MattPirkowski Yes, totally…and who has it proved useful to more times? That’s the source of this context — useful to who?

26 0
4/3/2025
Placeholder
Terminus Est@TerminusEst24• 8 months ago
Replying to @eshear

@eshear Counterpoint: there may be infinite patterns that start with 2, 4, 6, but there aren’t infinite patterns where all the information required to define the pattern is encoded twice in the 3-digit subsequence.

54 0
4/3/2025
Placeholder
Emmett Shear@eshear• 8 months ago
Replying to @TerminusEst24

@TerminusEst24 All the information to define the pattern is never encoded in the sequence

36 0
4/3/2025
Placeholder
Michael Fowlie@mwfowlie• 8 months ago
Replying to @eshear

@eshear Surprisingly, these tests are highly g loaded, so all smart people tend to get the same answer somehow.

2 0
4/4/2025
Placeholder
Emmett Shear@eshear• 8 months ago
Replying to @mwfowlie

@mwfowlie Yes, bc mind reading is a crucial skill that underlies all general intelligence. In a way, that’s all g really is.

2 0
4/4/2025
Placeholder
Polynumera@polynumera• 8 months ago
Replying to @eshear

@eshear This is a good observation about the futility of creating IQ tests for general intelligences. In particular I never liked ARC-AGI. But that doesn't mean they aren't reasonable measures for people.

0 0
4/4/2025
Placeholder
Emmett Shear@eshear• 8 months ago
Replying to @polynumera

@polynumera Never said they weren’t reasonable. I think they’re reasonable for AIs too!

2 0
4/4/2025