Shiva's Right Foot

• Shiva's Right Foot 13 Feb 2026 0:50 UTC

  2 points

  −3

  on: mod­els have some pretty funny at­trac­tor states

  I have a theory about the attractor states: they are in the center of the conceptual manifold.

  I’ve made an embedding of GPT2 and it seems to have a priciple gradient between “narrative” and “personal” language. On one end neurons that fire in discussions using “narrative” language, or impersonal topics frequently using passive voice. I think Wikipedia anchors this end of the manifold although news articles also have this passive, formal tone. The opposite end is “personal” language like Reddit interpersonal drama stories or calls to action in sales or politics; text where the writer or reader is an active participant (and some fiction written from first person perspective). It was suggested to me today that this somewhat mirrors the “assistant axis” of personality types recently discovered by Anthropic in that recent paper (the “assistant” end consists of personalities which would engage in detached impersonal statements of fact like “Evaluator” or “Librarian” while the opposite end has more personal personae like “Ghost” or “Narcissist,” notably “Visionary” seems near the center mass in one labeled figure).

  It occurred to me when looking at some points near the center that flowery biblical language or personal philosophy is basically the intersection of universal “narrative” style knowledge and the personal. The Bible is full of both narrative descriptions of historical events as well as things like lists of laws and procedures for ancient rituals but also has dramatic personal stories and makes direct appeals to the reader. The age of religious texts also suggests they have this basal quality which mixes many parts of language that gradually became distinct niches as language grew. Of course flowery text about personal philosophy on subjects like consciousness certainly is adjacent to ancient religious text and also exists at the intersection of the personal and universal.

Poly­se­man­tic­ity is a Misnomer

• Shiva's Right Foot 11 Feb 2026 22:11 UTC

  5 points

  1

  on: Where Will Call Cen­ter Work­ers Go?

  I think this may be neglecting the role of off-shoring which is an important (technology facilitated) part of the call center story. These are the lowest-skill, lowest-trust jobs that will be most likely to be replaced earliest by LLMs; in current configurations these employees (well, technically contractors) may be able to operate the sales website for the customer or tell them to power cycle an electronic device but they have to kick any real decision up an authority tree which then more likely ends up in a US call center.

  Literally: many of these outsourced low-level jobs are simply a voice interface for the website for the elderly and others uncomfortable with working a website. LLM systems are perfect for this task. So it could be that the largest impact of LLMs on customer service roles is happening overseas.

  And outsourcing these low-trust, low-skill customer service functions has been eating away customer service roles noticably in the BLS statistics since before the LLM boom started in 2022. Customer service roles have been one of the more rapidly shrinking job categories for years at this point in the Occupational Employment and Wage Statistics tables from BLS.

• Shiva's Right Foot 11 Feb 2026 16:22 UTC

  3 points

  −1

  on: Jevons Burnout

  In a way it is already here given how the anthropologists tell us participants in hunter-gatherer societies spend much more time on leisure activity than participants in industrial societies.

• Shiva's Right Foot 3 Feb 2026 20:28 UTC

  1 point

  0

  on: Shiva’s Right Foot’s Shortform

  GPT2′s “devotion” neuron may be pertinent to safety issues. I’m exploring GPT2-XL in OpenAI’s neuron viewer and 23:5243 is a neuron that basically captures the actions and hopes a person would have toward a loved one, likely a spouse. Activation on phrases like “stand up for,” “look out for,” “protect and enable,” “give comfort to,” and individual words like “strengthen,” “validating,” “reward,” or “benefit” seem to capture the ideal relationship between two mutually respecting individuals.

  Particularly the incorporation of the word “enable” along with “protect” seems to be a hopeful indication that the machine has a human-like wholistic conception of this idea rather than an unbalanced over-protectiveness as feared in some pessimistic scenarios. A neuron like this also may play some role in sycophancy; validating and strengthening the user is pretty precisely sycophancy behavior.

  https://​​openaipublic.blob.core.windows.net/​​neuron-explainer/​​neuron-viewer/​​index.html#/​​layers/​​23/​​neurons/​​5243

  This is part of a project where I am using a low dimensional embedding of GPT2-XL’s 24th layer to assist in interpreting the neurons. The embedding aids in interpretation for more difficult neurons, but this one was pretty straightforward. That said, the embedding does make me suspect this neuron is at least partially capturing romantic relationships because it is near to the erotic story neurons I discussed in a quick-take a couple days ago.

  I am trying to gauge interest in a longer frontpage post which will discuss larger findings and methodological details so any encouragement or expression of interest would be appreciated.

• Shiva's Right Foot 2 Feb 2026 16:05 UTC

  3 points

  0

  on: Shiva’s Right Foot’s Shortform

  One of the more interesting GPT2-XL neurons is this one for magic and fantasy. I’m exploring GPT2-XL layer 24 neurons using OpenAI’s neuron viewer and a low-dimensional embedding as a “map” of the conceptual space. 23:2721 is an interesting neuron that seems to trigger for situations involving fictional supernatural powers, particularly flavored for fantasy. “Lose yourself in a tale of magic and wonder,” is a good example of a high activation phrase; some directly mention figures from historical myth like “Grendel and his mom come busting in, forcing Beowulf to retaliate. Odysseus is just trying to get home,”

  https://​​openaipublic.blob.core.windows.net/​​neuron-explainer/​​neuron-viewer/​​index.html#/​​layers/​​23/​​neurons/​​2721

  The top activation phrase for 23:2721 seems to be a discussion of the troll in Ernest: Scared Stupid which I find pretty funny.

  The embedding I am using helps inform interpretation on nearby neurons in the embedding space. 23:2721 was near 23:1714 which discusses less fantastical fictional properties like Lena Dunham’s Girls or The Big Lebowski and neuron 23:6163 which discusses real historical events and figures like The Mayflower or Eisenhower.

  https://​​openaipublic.blob.core.windows.net/​​neuron-explainer/​​neuron-viewer/​​index.html#/​​layers/​​23/​​neurons/​​1714

  https://​​openaipublic.blob.core.windows.net/​​neuron-explainer/​​neuron-viewer/​​index.html#/​​layers/​​23/​​neurons/​​6163

• Shiva's Right Foot 1 Feb 2026 17:43 UTC

  7 points

  0

  on: Shiva’s Right Foot’s Shortform

  The funniest GPT2 neuron I’ve found: I am presently exploring GPT2-XL layer 24 neurons on OpenAI’s neuron viewer for an interpretability project. Last night I found a particularly funny neuron, 23:3024 (note layer indexing starts at 0).

  It activates on phrases about cooking instructions like “whip the double cream, vanilla seeds and icing sugar together” and phrases from erotic stories like “He wiped an abundance of Trent’s still-warm sperm off,” It is literally a food-porn neuron.

  I believe the concepts are united primarily through the word “cream” as you may note from the high activation phrases I just quoted. It is like quantifiable evidence that thoughts about food and thoughts about sex are highly intertwined in a generalized hedonistic concept.

  https://​​openaipublic.blob.core.windows.net/​​neuron-explainer/​​neuron-viewer/​​index.html#/​​layers/​​23/​​neurons/​​3024

  You need to click the “show more” for the high activations to see the erotic story segments; the topmost are mostly cooking instruction related. Incidentally, in the low-dimensional embedding I made there is another erotic stories neuron nearby in the embedding space. If you look carefully they are so similar you may notice there is even a repeated high activation phrase between these two on neuron viewer.

  https://​​openaipublic.blob.core.windows.net/​​neuron-explainer/​​neuron-viewer/​​index.html#/​​layers/​​23/​​neurons/​​4120

  I am considering a longer write-up of my neuron exploration project. I made a similarity embedding for the deepest layer of neurons which I am “walking” through in order to aid interpretability. With the embedding an investigator can see the embedded space area around the neuron and get clues from easier to interpret neurons for a difficult one.

  These two erotic-stories neurons are a good example of how the similarity space can inform interpretation. 4120 is more fully about erotic stories and the polysemanticity of 3024 may have thrown an investigator off (possibly deciding the neuron was only about cooking with some weird unexplained aspect) if the erotic content wasn’t confirmed by 4120 being nearby. Ultimately I am working on developing an intuition about the directionality in the embedding; I want to know the shape of conceptual space.

  It is fairly grueling work. Any encouragement or signs of interest would be appreciated.

  What links here?

  • Poly­se­man­tic­ity is a Misnomer by Shiva's Right Foot (12 Feb 2026 17:22 UTC; 11 points)

Dist­in­guish­ing Sen­sory Qualia Us­ing Neu­ral Structure

• Shiva's Right Foot 6 Jan 2026 22:22 UTC

  0 points

  −4

  on: On Own­ing Galaxies

  JM Smith makes a convincing case that property rights arise from an evolutionary game theoretic strategy which outcompetes both pure Hawk and pure Dove in the Hawk-Dove game. You even allude to the pervasive phenomenon of territoriality in animals:

  Dogs pee on trees to mark their territory; humans don’t respect that.

  Dogs do make marks and have surprisingly rigid and well defined territories. In as much as territorial rights to real estate constitute “property” it does not seem to depend on a human substrate (and rights to galaxies seem to fall within the conceptual boundaries of real estate territorial rights).

  The internet famous map of wolf-pack territories derived from location tracking collars:

  https://​​www.reddit.com/​​r/​​MapPorn/​​comments/​​cynexz/​​the_wolf_pack_map/​​

  For those unfamiliar it shows a surprisingly strict respect for boundaries; colorful wild scribbling trails which halt at distinct boundaries and do not cross over into the neighboring color’s scribbling trails.

  Smith shows that a set of players which use a coordination mechanism to decide which plays Hawk and which plays Dove will out compete pure Hawk or pure Dove players. His example is temporal precedence; i.e. players decide who owns the prize by who got to it first, much like how seating territoriality manifests for humans in something like a cafeteria or airport lounge. This is the root of animal territoriality as well.

  You can possibly make arguments that SAI will not use the same coordination mechanisms we currently use in terms of the entire extended legal apparati which express the various kinds of property rights in our society. However it does seem reasonable that property rights are preserved between asymmetrically powerful entities at present (like a middle class individual suing a large corporation) for similar reasons we should expect that property rights will continue: arbitrarily breaking the usual coordination methods for territoriality will harm the effectiveness of the coordination mechanism for less asymmetrical entity pairs; i.e. SAI stomping on human property rights may abrogate the system of property rights among SAI or, perhaps less catastrophically, coordinate other SAI on removing the property rights of the violator.

Well, Sea­sons Great­ings Every­one! [Short Fic­tion]

Shiva’s Right Foot’s Shortform

• Shiva's Right Foot 27 Nov 2025 15:48 UTC

  4 points

  0

  on: Shiva’s Right Foot’s Shortform

  Reddit Mod is a “job” AI will likely replace.

  There currently is a kerfuffle over the actions of a moderator of r/​art which seem to be a shining example of the petty and arbitrary cruelty in which Reddit mods are free to indulge as unpaid volunteers. This made me think that Reddit Mods are very similar to Dungeon Masters in Hasbro’s Dungeons and Dragons game, although why may not be clear to you readers.

  Hasbro’s DnD seems to be on a mission to alienate DMs from DnD in order to more directly access the players, which Hasbro llikely views as their real customers. It sees DMs as unpaid volunteers which intermediate the use of their product by consumers. In the past (under TSR) this was seen as a beneficial use of free labor but under the corporate ownership of Hasbro DMs were seen as extracting value from the role of player intermediary which could theoretically go to Hasbro’s bottom line. Hasbro until recently had been pursuing an online strategy which would move the pen-and=paper RPG into an online platform which likely eventually would obviate the need for DMs. In an online platform directly controlled by Hasbro not a single dollar of Hasbro’s money would be lost to a DM’s horny depiction of a female character, the DM’s unique approach to personal hygiene, nor the arbitrary and unfair cruelty of a power=tripping DM.

  Outside of the online platform Hasbro has also moved to shift power away from DMs and towards players. Mechanically there are things like the new death save system which makes a player character death much more mechanically difficult than in past editions of the game. The caveats in old editions that the written rules are just suggestions and that the rulings of the DM are final have been removed. Recently there has been rumors of a move to disallow the DM screen, a folding barrier which allowed the DM to conceal things like maps of areas yet unseen by players or detailed character sheets for NPCs. All of this seems designed to dimish the role of the DM.

  And a similar effort to more directly control the expereince of end-users likely will happen at Reddit. This latest controversy is over a mod not only permanently banning a professional artist from r/​art for an offhand comment that prints of the art he posted were availible but also erasing his entire r/​art post history simply for offering an apology and asking to be reinstated. This is pretty representative of the arbitrary nature of Reddit mod power and likely has been repeated many times in cases where the user did not have access to a large audience on a different platform, as did the artist in this case.

  Reddit will likely recognize that an AI facilitated system could let them hire a handful of paid moderators who will be committed to act in a fair and professional manner, disintermediating the existing volunteer moderators. Reddit is already a bulk of the training data for most LLMs and Reddit itself already has a massive dataset of moderator actions which could be used for training. All it would likely take at this point is a major controversy or perhaps a string of moderate controversies to motivate them to implement a new system.

• Shiva's Right Foot 23 Oct 2025 16:15 UTC

  1 point

  0

  in reply to: James Diacoumis’s comment on: Con­tra-Zom­bies? Con­tra-Zom­bies!: Chalmers as a par­allel to Hume

  I had a very busy IRL day yesterday and have intended to respond to this.

  While I am initially inclined to simply do what you ask out of kindness I am still convinced that I have no real reason to do so and therefore acceding here may portray me as a pushover. This really is an instance where some human neutral third party input to this dispute would be extremely helpful and I wish there was more of a culture online of such interventions. I would expect there to be such a culture here on Lesswrong, but perhaps not.

  Nevertheless I did consult a non-human mediator. I prompt Gemini with the following in a new chat:

  I wrote a post where I said Chalmers’s p-zombie argument is so bad “How could any rational person accept something so absurd?” but then go on to say that his argument is more persuasive from color inversion. Someone else characterized this as arguing against Chalmers. Is that an accurate characterization or should they retract it?

  As a guard against sycophancy I do phrase it from your perspective. As far as I know any personalization feature of Gemini is turned off, including remembering past chats. Even if I’ve incorrectly assessed Gemini’s outside knowledge, previously I did not interact with Gemini about this post, these comments, nor the general idea of people on Lesswrong or in the Rationalist community arguing against Chalmers.

  Its response concludes:

  ##Conclusion on the characterization

  The other person’s characterization that you are arguing against Chalmers is accurate with respect to the p-zombie argument. You explicitly argued against the p-zombie argument’s soundness and persuasiveness.

  They do not need to retract the statement that you argued against him, as you undeniably argued against one of his main arguments. You simply qualified your opposition by showing that you find a related argument to be more compelling. You are essentially a selective critic of his methods, not a complete opponent of his conclusions.

  https://​​gemini.google.com/​​share/​​52f7996e8472

  This is an increasingly interesting intellectual exercise and I would be open to alternative prompting techniques or perhaps the opinions of other LLM systems as a comparison.

• Shiva's Right Foot 21 Oct 2025 14:53 UTC

  1 point

  0

  in reply to: James Diacoumis’s comment on: Con­tra-Zom­bies? Con­tra-Zom­bies!: Chalmers as a par­allel to Hume

  You linked my post Beyond the Zombie Argument as evidence of people on LessWrong purportedly arguing against Chalmers which is a pretty significant misreading of my article.

  Your article starts with “As a staunch physicalist I’ve spent a long time wondering how anyone could possibly find the zombie argument compelling.” and proceeds to address Chalmers (1996) at one point asking “How could any rational person accept something so absurd?” This seems to purport to argue against Chalmers.

  You do seem to say that the color inversion argument is a more convincing argument. However, the disagreement with the zombie argument I think does carry my characterization. Additionally your post somewhat misses Chalmers’s eventual dismissal of these questions (well, dismissal in the way Hume dismisses non-empirical questions by placing them outside of scientific inquiry, which arguably is not total dismissal).

  I really don’t mean any disrespect by this.

Con­tra-Zom­bies? Con­tra-Zom­bies!: Chalmers as a par­allel to Hume

• Shiva's Right Foot 13 Oct 2025 7:10 UTC

  3 points

  0

  on: The Most Com­mon Bad Ar­gu­ment In Th­ese Parts

  This does seem to be the form of argument used when it is demanded that someone proves a negative. I’m surprised no one has brought that up elsewhere in the comments.

  It is incumbent on someone making a positive statement to provide evidence. When someone doesn’t provide enough evidence for a positive claim critics will point that out but “there is insufficient evidence” can only be repeated so many times.

  People continue to press “Well, prove why [positive claim] isn’t the case!” and the critic’s only response is to iterate through all the positive arguments made for [positive claim] and show they are insufficient.

• Shiva's Right Foot 24 Sep 2025 23:16 UTC

  0 points

  0

  in reply to: dsj’s comment on: A Com­pat­i­bil­ist Defi­ni­tion of Santa Claus

  We do not have to meticulously caveat that when we discuss the concept of “night” we are not talking about the phenomenon of the Sun moving away from our position on the Earth disc to a point where mountains cast our current position in shadow but rather the phenomenon of the Earth rotating away from facing our current position towards the Sun in order to avoid “lying” to any Flat-Earth believers who are listening. In fact I don’t think discussions arranging a meeting with a group of Flat Earthers “tomorrow night” would require any such caveats or explanations nor incur any case of “lying,” despite what are presumably somewhat different conceptions of what “night” is. Such minor differences are for the purposes of such a conversation basically completely negligible.

• Shiva's Right Foot 24 Sep 2025 16:30 UTC

  1 point

  −5

  in reply to: dsj’s comment on: A Com­pat­i­bil­ist Defi­ni­tion of Santa Claus

  very different

  The degree to which his definition is “very different” is not clear. Definitions vary at least slightly from person to person all the time but we don’t make long semantic declarations in normal conversation unless it serves some specific functional purpose.

• Shiva's Right Foot 24 Sep 2025 14:01 UTC

  1 point

  0

  in reply to: dsj’s comment on: A Com­pat­i­bil­ist Defi­ni­tion of Santa Claus

  What purpose is the new definition of the term “Santa Claus” serving here

  The child can engage his friends in conversations about what gifts Santa Claus brought him.

• Shiva's Right Foot 24 Sep 2025 5:22 UTC

  2 points

  0

  in reply to: dsj’s comment on: A Com­pat­i­bil­ist Defi­ni­tion of Santa Claus

  In this case the analogy would be that compatibilists are parents explaining to confused children that the “Santa Claus” which is delivering presents to their school friends is in fact the generosity of those children’s parents. Perhaps some more religious parents would be defining Santa Claus or Free Will with reference to more fantastical elements such as a man in a red suit or an extradimensional ethereal substance which somehow telekinetically manipulates our bodies.