If the most powerful language-processing system we’ve ever built encodes ideology as a separable geometric feature — not as content but as orientation — then maybe the reason Occupy failed is that...
Just about any concept that could be mapped out along a range of values should be representable as a n dimensional vector in language space. It is a good idea to map out for each popular models the values of the built in bias for a set of such concepts. There is nothing mystical about this. In terms of persuading humans. It is unclear how hidden vectors of bias can be changed except through first level language constructs given the limitations of human senses.
Let me say if I understand your research. Basically, just introducing a vector, without changing the weights or anything, the model behavior changes. Then you point vaguely to spiritual traditions. I think this is a very interesting concept and I have two questions. Have you check if any development in the model can make him resilient to that vector, that is for resistance, and what would be in humans that vector? Values, spiritual beliefs? But that cannot be because it would be part of the weights. I am an academic philosopher in philosophy of science and pragmatism and also I have 30 years of practice in the esoteric and spiritual world. Your research is stunning because I have the same question. How do you change society so we don't end up killing ourselves? I would be interested in discussing your ideas if you want. Maybe our combined perspective can give us some insights.
Lovely article Micah. Just downloaded and checking out OutCry, well done. Graeber was, and remains, my inspiration. It sucks how movements can win the discourse while still failing to shift the deeper orientation of power.
That’s why we started Common Planet Foundation: to move beyond critique and persuasion into new civic/economic infrastructure. We’re building Creditism and the AYU Network as a way to reorient credit, commons, and governance around life rather than debt, extraction, and survival.
If activist tech needs a new compass, economic architecture has to be part of that geometry too.
Regarding this being an important aspect of LLMs, yes. Regarding human minds, I think probably no. An LLM type of "generative AI" seems to be eerily similar, functionally, to human sensory memory recall. That's it. Nothing about the rest of the brain function, no thinking. We have to grow these things to hold most of the written word and then chain and loop them to get anything resembling more, and that's not what a brain, largely, is.
That’s the big question: how much can be generalized from LLMs to humans? The more I dig into LLMs the more I think they unlock a new understanding of humans.
Supernatural doesn't exist. All is part of nature. Even quantum non locality of the mind. Once the Materialists finally accept quantum mechanics into their fold (and therefore Psi effect) we will finally move forward. The Elites spent 500 years with lying to the masses that only atoms and molecules matter (all the while all Enlightenment philosophers and 19th aristocratic scientist were part of Elite Occult groups). They hoodwinked the masses that "magic" aka Gnosis doesn't exist because Gnosis would spead uncontrollably because it is information and ultimate form of power/equalizer. Capital physical machinery and material stuff? That's easy to control with private property laws and pigs in uniforms. Information once out cannot be contained.
The AI Kabbalist got his algebraic wings
Just about any concept that could be mapped out along a range of values should be representable as a n dimensional vector in language space. It is a good idea to map out for each popular models the values of the built in bias for a set of such concepts. There is nothing mystical about this. In terms of persuading humans. It is unclear how hidden vectors of bias can be changed except through first level language constructs given the limitations of human senses.
CAA steering to adjust bias
Have you mapped the relationship between the model and how it relates to human behavior yet?
Interesting question. Say more
Let me say if I understand your research. Basically, just introducing a vector, without changing the weights or anything, the model behavior changes. Then you point vaguely to spiritual traditions. I think this is a very interesting concept and I have two questions. Have you check if any development in the model can make him resilient to that vector, that is for resistance, and what would be in humans that vector? Values, spiritual beliefs? But that cannot be because it would be part of the weights. I am an academic philosopher in philosophy of science and pragmatism and also I have 30 years of practice in the esoteric and spiritual world. Your research is stunning because I have the same question. How do you change society so we don't end up killing ourselves? I would be interested in discussing your ideas if you want. Maybe our combined perspective can give us some insights.
good question, and count me in for that discussion. Would love to compare notes. My team and I just launched: https://common-planet.org/
Lovely article Micah. Just downloaded and checking out OutCry, well done. Graeber was, and remains, my inspiration. It sucks how movements can win the discourse while still failing to shift the deeper orientation of power.
That’s why we started Common Planet Foundation: to move beyond critique and persuasion into new civic/economic infrastructure. We’re building Creditism and the AYU Network as a way to reorient credit, commons, and governance around life rather than debt, extraction, and survival.
If activist tech needs a new compass, economic architecture has to be part of that geometry too.
Would love to connect and compare notes:, just launched https://common-planet.org/
Regarding this being an important aspect of LLMs, yes. Regarding human minds, I think probably no. An LLM type of "generative AI" seems to be eerily similar, functionally, to human sensory memory recall. That's it. Nothing about the rest of the brain function, no thinking. We have to grow these things to hold most of the written word and then chain and loop them to get anything resembling more, and that's not what a brain, largely, is.
That’s the big question: how much can be generalized from LLMs to humans? The more I dig into LLMs the more I think they unlock a new understanding of humans.
Supernatural doesn't exist. All is part of nature. Even quantum non locality of the mind. Once the Materialists finally accept quantum mechanics into their fold (and therefore Psi effect) we will finally move forward. The Elites spent 500 years with lying to the masses that only atoms and molecules matter (all the while all Enlightenment philosophers and 19th aristocratic scientist were part of Elite Occult groups). They hoodwinked the masses that "magic" aka Gnosis doesn't exist because Gnosis would spead uncontrollably because it is information and ultimate form of power/equalizer. Capital physical machinery and material stuff? That's easy to control with private property laws and pigs in uniforms. Information once out cannot be contained.