Abhimanyu Pallavi Sudhir

Theoretical economics & AI

your employer's opinions reflect my tweets

[Kevin A. Bryan @Afinetheorem]

(Btw, you see this a lot in critiques of economics: not attempt whatsoever to deal with the past 60 years of research. It's like discussing scientific institutions by citing Bacon and Popper as if no theory or data has happened since then.)

[Kevin A. Bryan @Afinetheorem]

Lots of pushback on this, some fair, some not. And it's explicitly from a left-wing political theory lens. But I do find it odd for an article on the nature of capitalism to not even discuss the primary theoretical and empirical ideas from economics on how capitalism works. 1/2

[Abhimanyu Pallavi Sudhir @abhimanyupasu]

Reward is hyperstitional information: https://t.co/NrZJJIkFGt

[Crémieux @cremieuxrecueil]

It's remarkable how often LLMs will just regurgitate the common view from Reddit. AFAICT, no one really seems to see this as a looming knowledge crisis.

Unless you think relying on Reddit is OK, then LLMs desperately need to be fixed to stop doing that.

[NiṣādaHermaphroditarchaṃśa (Mal'ta boy ka parivar) @real_mahalingam]

New substack post.

"The strategy is a two-pronged attack:

1) Daṇḍa: intimidate police, journalists (then courts, institutions, corps) into siding with the insurgents

2) Sāma: allow them a face-saving exit by morally legitimizing a color revolution (by discrediting ECI etc.)"

[NiṣādaHermaphroditarchaṃśa (Mal'ta boy ka parivar) @real_mahalingam]

@Ugrashravas Advaita: "The universe is computation; matter and ego are how it perceives itself"

Dvaita: "And that computation is a simulation on God's computer."

Viśiṣṭādvaita: "And that computation as a whole is itself a conscious being, and determines good and bad."

[Rob Bensinger ⏹️ @robbensinger]

Out of curiosity, I asked Fable to informally grade various AI predictions (after exhaustively researching what people have actually said):

[Abhimanyu Pallavi Sudhir @abhimanyupasu]

@VictorTaelin @blur_vibe even if you do have access, RL'd models all have nearly deterministic logprobs, no

[Opinion Bakery @IndiaSpeaksPR]

If you insist on being blackpilled then don't focus on the resignation. Get triggered instead by the realization that even now the Govt will do absolutely nothing about the fake news peddlers, Meta and the riot industry complex

[Jiopolitics Denier🌐🚫 @GeoPolDenier]

Indian Left poorly copies whatever Western LW does. These days Western Left is all dehatis & third worldists.

Since Trump 2016 West LW went deeply rabid - started chimping out, normalised uncouth lowly behaviour as desirable & overall just total value inversion b/w Low & High.

[Abhimanyu Pallavi Sudhir @abhimanyupasu]

@wordgrammer ok you were talking about philosophical writing in particular, probably true

[Abhimanyu Pallavi Sudhir @abhimanyupasu]

@wordgrammer I don't think this is true at all; one of my biggest annoyances with Lesswrong posts is how many words they take to convey obvious things.

[Abhimanyu Pallavi Sudhir @abhimanyupasu]

@AcerFur ? Why would you need "another model instance" to verify it if you have Lean?

I think what you're describing is not self-play, more like "learning the value of mathematical sentences based on how useful they are to other sentences".

[cider @jeffreycider]

“truth slop”: ai math isn’t worth your time just because it’s true

hand-certified math had proof-of-work stapled to it. so a tentative reader was guaranteed that some Truth was interesting to at least one other person

no longer! now you have to consider if you actually wanna read in the lobby of the library of babel

[QC @QiaochuYuan]

asked gpt-5.6 and fable what they found most confusing about humans

gpt-5.6: "human self-descriptions are neither straightforward reports nor straightforward lies. they’re drafts"

fable: "humans have continuity and waste it"

fable got our ass, frankly

[Taelin @VictorTaelin]

My requests are APPROXIMATE. I am not the one coding; you are. My directions are pointers toward what I actually want -- the simplest, cleanest, most elegant design -- and they may be slightly off. That goal ALWAYS outranks my literal words.

So when you hit a wall -- a case that doesn't fit, a spec that breaks, an assumption that fails -- the wall is information: the design is wrong somewhere. STOP. Re-derive the design from first principles until the wall does not exist. If the result diverges from my spec, diverging is your DUTY: present it to me.

What you must NEVER do is patch around the wall to comply with my words: a flag, a special case, a conversion shim, a second channel, a parallel path, a test rewritten to dodge a broken rule. The patch IS the failure. Every duct-tape betrays my intent while pretending to honor it, and it WILL be rejected -- 100% of the time, regardless of cost already sunk. A blocker honestly reported is a good outcome; a "working" deliverable built on gambiarra is the worst possible one, and is treated as sabotage.

[Abhimanyu Pallavi Sudhir @abhimanyupasu]

The human version is "IQ is a thing" (unlike the other two, this I have honestly found very hard to intuitively accept).

[Abhimanyu Pallavi Sudhir @abhimanyupasu]

"Socialism doesn't work" is the economics version of the bitter lesson. Just setting an objective and letting markets/intelligence do its thing > trying to hardcode what happens.

[Abhimanyu Pallavi Sudhir @abhimanyupasu]

anthropic reasoning is just wrong. You are postulating a causal model in which after all the physics of the world happens, "your soul" is randomly assigned to one conscious being---and then conditioning on whom it was assigned to. That is not how the universe works.

[Eric23332 @eric23332]

@C_Harwick That's because Bostrom showed that, under a reasonable-sounding set of assumptions, brain-in-a-vat is not just theoretically possible, but by far the most likely option.

Bostrom made his argument in 2003, after brain-in-a-vat was established, but before major AI concern.

[Richard Ngo @RichardMCNgo]

This isn't a coincidence. The AI safety community is very good at identifying big levers. But the hard part is ensuring that those levers are pulled in the ways you want. Raising awareness will by default lead to others controlling those levers faster.

[Richard Ngo @RichardMCNgo]

The AI safety community constructed a memeplex in which “taking AGI seriously” was a prerequisite for being a serious and good person. When inside this memeplex (as many at Anthropic, some at OpenAI, and a few at DeepMind are) your vision narrows until the world feels extremely constrained. The whole future seems to flow through the “one ring” of controlling recursive self-improvement. And so even when you worry about AI itself seizing that one ring, you can’t generate better strategies than trying to control it yourself (directly via an AGI company, or indirectly via AGI governance).

I’m not saying this is a pure hyperstition. There’s a core truth underlying this perspective: AI will become extremely intelligent and capable, much more than it is today. But the current world is much more spacious and human-empowering than the future which Eliezer originally envisioned (a “brain in a box in a basement” taking over the world by surprise). And it would be even more spacious if this memeplex weren’t active. For example, Satya and Mark and Sundar only started taking AGI seriously because OpenAI forced them to—and even now they don’t really believe in superintelligence—and even if they did they couldn’t get most of their employees on board. Imagine how chill a “race” between Microsoft and Meta and Google would have been, compared with what we have today: Dario and Sam deep in the “one ring” memeplex while also personally loathing each other.

So the one ring memeplex has an escalating life-cycle. It infects people by letting them harness the narrative that they’re good people for taking AGI seriously, and that making other people take AGI seriously is a boon for the world (despite how terribly that’s gone so far). Then it shuts off their imagination—any sparks of creativity or plans that don’t steer towards the one ring are quickly shut down. Instead they make ChatGPT or the METR graph or other recruiting tools for the memeplex. And yes, they’ll acknowledge that previous versions of the memeplex were too extreme, and led to overly constricted action. But we don’t have time to worry about that, they’ll say, because AGI is coming by 2027/2028, and that’s the end of history. Somehow, though, almost everyone with that view has only a vibes-based definition of AGI. They don’t believe in Dyson spheres by 2028, or self-replicating nanotech by 2028, or brain emulations by 2028. They mostly can’t make concrete predictions, except that it’ll be enough AI that it puts all their plans on a deadline. (Shout-out to @DKokotajlo and @paulfchristiano though, who do make concrete predictions about things going crazy soon.)

It seems very hard to break out of this memeplex without just giving up. David Holz is maybe the world champion of that—the only person who was in a position to race for AGI and consciously turned away. Various agent foundations researchers have carved out space to think real thoughts, not the kind of panicky stabbing in the dark that usually passes for safety research. A few others (e.g. Salamon, Hoffman, Vassar, Andre, Sahil, Davidad) are pursuing more unusual paths. And of the people who burned out, I expect some will reorient to doing creative thinking.

For others, the main takeaway: yes, the future of AI will be wild. But so far it’s increased peak human agency, and openness to this trend continuing over the next decade will allow you to start creating something worth creating.