Jerdle

Karma: 47

Jerdle 31 Mar 2026 16:24 UTC
10 points
2
on: Jerdle’s Shortform
What’s going on with interpretability these days?

I found the whole monosemantic sparse autoencoder idea interesting, but this was 2023 and it’s now 2026.

Jerdle 27 Mar 2026 17:20 UTC
1 point
0
on: Musical Outgroups
2026 update:

The Blue Tribe has remained roughly as extreme as in 2020, but the specific ideology has changed. Now, it’s all about Palestine.

The Red Tribe has continued along the same path as between 2017 and 2020.

The Grey Tribe has come into full fruition, and has spread out if not entirely split, mimicking the wider Blue-Grey-Red split within itself.

Some Greys are YIMBY abundance liberals, merging with the Purples.

Other Greys (who I think of as the core Greys) are the direct outgroup of the Blues. NFTs in 2020-21, AI from 2022-25 and prediction markets in 2026 have all been opposed by the Blues, and are closely correlated because they’re all Grey Tribe.

And finally, there’s the Greys that have allied with the Reds. Musk and Yarvin come to mind here.

I don’t actually know what happened to the Pinks.

Jerdle 26 Nov 2025 0:23 UTC
1 point
0
in reply to: Raymond Douglas’s comment on: mattmacdermott’s Shortform
Was the sixth one a blank canvas?

(This is only going to make sense to a fairly small audience)

Jerdle 20 Nov 2025 10:42 UTC
1 point
−3
on: Status Is The Game Of The Losers’ Bracket
I know you’re using “loser” in a non-Raoian sense, but those levels map really well onto the loser/clueless/sociopath trichotomy.

Level 0: Loser Level 1: Clueless Level 2: Sociopath

Jerdle 20 Nov 2025 10:38 UTC
3 points
0
in reply to: Portia’s comment on: I Am Scared of Posting Negative Takes About Bing’s AI
Future here: Yes. Yes we did see that. GPT-4o was a nightmare for it.

Jerdle 7 Aug 2025 18:31 UTC
7 points
2
on: When will GPT-5 come out? Prediction markets vs. Extrapolation
As it happened, it was August 7, 2025. Update your models accordingly.

Jerdle 7 Apr 2025 22:25 UTC
1 point
0
on: PauseAI and E/Acc Should Switch Sides
I get that this was a joke, and don’t agree with the “AGI doomers should push for faster development” part, but I actually do agree that e/acc would be better off accepting a slowdown now in order to get more powerful AI later.

Jerdle 9 Mar 2025 11:08 UTC
1 point
0
in reply to: Gurkenglas’s comment on: Jerdle’s Shortform
It could do, but a represents the amount of utility remaining.

Maybe the more natural thing would be to have a be the effective tax rate, and have it be (z/x)^a.

Jerdle 5 Mar 2025 22:02 UTC
1 point
0
in reply to: Gurkenglas’s comment on: Jerdle’s Shortform
x is the initial income, and I forgot to cancel it. Good point.

Turns out, it’s far simpler than I had it as.

Jerdle 5 Mar 2025 21:04 UTC
3 points
0
on: Jerdle’s Shortform
Pseudo-flat tax formula:
Assume utility is logarithmic in income, and the goal is to set the experienced tax burden to be constant.
Then, we have the formula that the average tax rate, where $a$ is a parameter controlling the experienced tax burden and $z$ is the break-even point, is as follows:
$f (x) = 1 - {(\frac{x}{z})}^{a - 1}$
$x$ is the input income, and $f (x)$ is the average tax rate.

Jerdle 24 Feb 2025 1:10 UTC
1 point
0
in reply to: bokov’s comment on: The Strangest Thing An AI Could Tell You
My assumption here would be, not that everyone I knew was evil, but that I was evil and was being told this as a level-2 means of getting me away from potential victims.

Jerdle 10 Feb 2025 0:58 UTC
2 points
0
in reply to: CapResearcher’s comment on: The Outer Levels
Thanks for the review!

I will start by saying that the odds and evens structure isn’t especially original. I’ve seen similar things discussed in previous posts and comments, mostly as a dimension of a 2x2 grid. As such, your review seems like the classic quote, falsely attributed to Samuel Johnson: “this is both original and good, but the original parts aren’t good and the good parts aren’t original”.

Of course, I disagree with that (if I thought the new levels weren’t useful, I wouldn’t have come up with them), and will try to explain why the original parts are good, or at least have some value.

To start with, level −1. This is pretty useless. I’m not sure it actually even exists, it’s just a natural consequence of the structure. Inasmuch as it does exist, it is incompatible with the existence of an agent. Perhaps the view of a camera would be level −1.

However, I see levels 5 and 6 as real, useful, and importantly not the same as levels 3 and 4.

Specifically, I see levels 5 and 6 as ungrounded. Levels 3 and 4 are not grounded in object-level reality, but are at least grounded in something, namely signalling and tribal affiliation. While “there’s a lion across the river” no longer means anything about actual lions, it still means something. It could not be replaced with “there’s a foobar across the bazquux”.

At the recursive tier, the words “lion” and “river” become irrelevant, and the system of references no longer roots itself in reality. Now this could presumably be considered “mask[ing] the absence of a basic reality”, but I’m not sure it even masks it. Reality is just absent. This is Baudrillard’s true level 4 of “pure simulacrum”.

Part of me wants to renumber the entire system, given that this is allegedly Baudrillard’s model. In this case, the political tier would all be level 3 and the recursive tier would all be level 4.

Jerdle 4 Feb 2025 22:15 UTC
1 point
0
in reply to: hamiltonianurst’s comment on: Pick two: concise, comprehensive, or clear rules
What, Dimir Bastard?

Jerdle 4 Feb 2025 19:52 UTC
1 point
0
on: The Outer Levels
I’m not entirely sure my use of that example as level 5 was accurate, it might actually be levels 3 and 6, because the goal of the slogan is to change people’s maps to include a recursive statement rather than merely to express a belief in a recursive statement.

Jerdle 4 Feb 2025 15:27 UTC
1 point
0
on: Newcomb’s Problem as an Iterated Prisoner’s Dilemma
I have found a separate post that firms up the dodgy second paragraph. I would have linked to it, but while there’s a time loop involved in the theory, there wasn’t enough of one to link to a post written in 2024 in my post in 2022.
In Strategic Time, Open-Source Games Are Loopy

Jerdle 4 Feb 2025 14:29 UTC
1 point
0
on: The Self-Reference Trap in Mathematics
This is very closely related to a discussion in my post on simulacra levels, but I think this is importantly false.
Yes, mathematics refers to nothing outside itself, but that is not to say that things within it cannot prove other things within it. It is a highly abstract field (my simulacra level 5, which is a form of Baudrillard’s 4), but that is not why it hasn’t been solved.
Take the question “what is 2 + 2”. This is made up of pure abstraction (calling this abstraction simply “1s” is not actually correct—“1” is a quantity defined in this abstraction), and does not directly apply to reality, except through a translation layer where the symbol “2″ is equated to * *. However, the answer is still 4.

Jerdle 9 Sep 2024 2:49 UTC
4 points
3
on: In Strategic Time, Open-Source Games Are Loopy
It’s a shame I wasn’t able to use a time loop to have this exist when I wrote my post on Newcomb’s problem being an iterated PD.

This concept is exactly what I needed, and it’s explained far better than I did.

Jerdle 15 Aug 2023 15:17 UTC
1 point
0
in reply to: Jerdle’s comment on: PD-alikes in two dimensions
Expanding on the Y>W=Z>X and X>W=Z>Y, I would split Abundant Commons at Y=Z, into Abundant Commons above the line and Deadlock below it. Then, the games equated are Deadlock and the PD, and those form a natural continuum.

Jerdle 23 Jun 2023 10:40 UTC
3 points
0
on: PD-alikes in two dimensions
Loosely and non-rigorously, x/0 is infinite, and so all games with W=Z are extreme forms of the corner games (unless X=W=Z or Y=W=Z). X~Y>W=Z gets you an anti-coordination game and W=Z>X~Y gets you a pure or relatively pure coordination game (Let’s Party).

Y>W=Z>X and X>W=Z>Y are interesting, because they equate games as different as the PD (or Too Many Cooks) and Abundant Commons. I would describe this game as more similar to the Abundant Commons than the PD, as Flitz/Flitz is a perfectly acceptable equilibrium. The value transfer here is neither hyperefficient nor inefficient, but merely efficient.

The triple equalities here are equivalent under name change, so, WLOG, let’s take X=W=Z. Then, there are two games: Y>X and Y<X. Looking at the diagram, X=W=Z>Y should resemble Studying for a Test, while Y>X=W=Z should resemble the Farmer’s Dilemma.

The former game has a primary theme of avoiding Y, and so, while Flitz/Flitz is an equilibrium, I would expect to see more Krump/Krump, as it is never beneficial to play Flitz when there’s any risk of Krump.

The latter game is more complex, but the equilibrium you actually see is Flitz/Flitz, because the only way to get Y is if you play Flitz.

Finally, with all four equal, there is no longer much of a game. All strategies are equilibria, the payoff is identical in each case. This is the trivial game.

Jerdle 14 Oct 2022 23:05 UTC
−3 points
0
on: Why I think nuclear war triggered by Russian tactical nukes in Ukraine is unlikely
I think your 5% figure of KABOOM given retaliation fails to condition on kaboom.

I would estimate an 8% chance of kaboom (broadly following prediction markets, going by a 10% chance of the order to nuke Ukraine and an 80% chance of it actually happening) and an 80% chance of retaliation. For KABOOM, either the West or Russia would drop the first strategic nuke. For the West, the probability is somewhere around 0.1% given kaboom. For Russia, while the probability of KABOOM given escalation is probably less than 5% (more like 1%, I’d guess), kaboom has happened, and so the probability of Russia dropping a strategic nuke is closer to 12.5%. And so, the final probability of KABOOM is closer to 0.9% than 0.25%.