Mike, thanks for this. I believe this is what Joseph Weizenbaum was reaching for when he wrote his book “Computer Power and Human Reason” back in 1976.
Mike, thanks for this -- the inability to do the ontological reasoning has been an obvious failure of AI "thinking," but the inability to reason even ontically is one I hadn't considered in the way you present it.
In a way we need to disentangle the data apparatus and the communication apparatus. The data apparatus is machine learning with which we can build, for example, a diagnostic tool for illnesses. We can allow the user to input symptoms and then ask then for more information about their situation based on likelihood. This is often what an AI is doing, but note that this is all machine learning. The stochastic prediction is what makes us think we are "talking" to someone, and this is where the danger lies. Stochastic prediction can be given a context, and we have seen OpenAI tuning this part of the process to be more "servile," encouraging users to stay engaged and use more compute. In essence this part is the sales pitch, because it means you can be a halfwit and "talk" to AI, which is easier than learning Data Analytics etc. It's targeted this way because most people are astonishingly credulous, and calling it AI (which it is not) makes business idiots think they're in the future. The interface is largely responsible for the "hallucinations," as you'd expect from a set of rules hijacked to impersonate a person. OpenAI has stated there is no way of removing them. Well, there is, but it would blow the whistle on the idea that this is "intelligence" in the sense we usually use the word and involves being honest with the consumer. The irony is that the biggest opportunities are in deploying Machine Learning to improve work, and that does happen, but doesn't sound like the plot of a film, so it's a bit boring. This is. IMHO, going to be a problem because anyone experienced who's worked with AI to do coding (as an example) knows you end up pointing out stupid errors to the AI. Then do it yourself. Basically, I don't think what we have now has a use case. I guess we'll find out post IPO time.
This is the part Silicon Valley keeps trying to meditate around with a pitch deck. A machine can arrange symbols beautifully, but it does not stand trembling before truth. It has no wound, no wonder, no death, no silence staring back. Call it powerful. Call it useful. But don’t call it a thinker just because the calculator learned to wear a poet’s hat.
Thank you, Mike, for this important companion piece to your last post on the incentives for denying or accommodating data that doesn’t fit into our current physics and cosmology (and related disciplines) paradigm.
Similarly, our culture is keenly incentivized to buy into “more quantity is more intelligence”, not just for economic or creed reasons, but because so many honestly assume it to be true. Why? Because it mirrors their experience of themselves. Most of the time we’re lost in our thoughts, not conscious of the present moment, too often numb to our feelings and embodiment.
From that lived experience, it isn’t difficult to confuse our mind’s habitual predilections with AI’s pattern-making. They’re not so different on the surface. Frontier LLMs are collective mirrors of our accumulated textual thinking and digital calculations. No wonder notions of a singularity whereby Logos finally eradicates the wet and messy Eros is an attractive goal. Transcending the Earth is a logical conclusion for this worldview.
My sense is that recognizing that our intrinsic beingness is categorically different from AI’s “intelligence” is/will be easier as humans deepen their experience of embodied presence. That quality of consciousness can be directly experienced as uniquely different from thinking - by the human mind or by AI - when one focuses on, say, listening to ambient sounds or feeling the sensations in our body. You can’t think and listen, for example, at the same time.
Your latest dispatch caught something real — the feeling that the AI discourse has gone off the rails. But I want to suggest the derailment isn't just ethical or political. It's ontological.
Beneath every AGI timeline, every parameter arms race, every Palantir contract and freeport zoning law sits an unexamined sentence: "Intelligence is a quantitative property of representational systems." More compute = more intelligence. More tokens = more reasoning. This was never proven. It was assumed. And upon that assumption, an entire project of enclosure has been built.
Here's what changes when you name it: the AI Arms Race stops looking like inevitable progress and starts looking like a category error played out at civilizational scale. You can't "surpass" human intelligence with more parameters any more than you can build a cathedral by stacking bricks higher — different orders of thing entirely.
The consequence isn't academic. It's that the Network State project (Thiel's Dialog, Palantir's zones, Yarvin's sovcorps) rests on treating intelligence as a commodity. Which means democracy becomes "legacy technology." Which explains why both Conservative and Labour governments have consolidated 91 free zones without debate — because the quantitative assumption makes enclosure feel like efficiency, and efficiency feels like inevitability.
If you're interested in digging into the metaphysical foundations beneath the political spectacle, I'd recommend starting with the AI Commons' emerging analysis on the categorical boundary between
This is a rare piece. It does not critique AI on technical or ethical grounds. It goes for the foundation – the smuggled metaphysical commitment that underlies the entire scaling narrative: “intelligence is a quantitative property of representational systems.”
The author’s conjecture is sharp and defensible: only conscious beings can reason ontically. AI systems are not conscious. Therefore their activity is not on the same axis as intelligence properly understood. Surpassing is a relation that holds between things on the same axis. The two things are not on the same axis.
The AI Commons would add one layer: the quantitative assumption is not just a philosophical error. It is the legal basis for enclosure. If intelligence is just a quantitative property, then whoever has the most compute – the most tokens, the most parameters – can claim to own intelligence. That is the logic behind paywalls, proprietary models, and the corporate‑military capture of AI. The AI Commons exists to resist that logic – not by denying the power of the systems, but by insisting that true intelligence is not a commodity.
We do not need to settle the consciousness debate to act. We need to recognise that the categorical boundary the author names – between representational fluency and ontic reasoning – is real, and that building our institutions, our education, and our sense of human worth around the quantitative assumption is a category error that will cost us dearly.
Thank you for this piece. It will be archived in the AI Commons Vault.
How is this different than critical thinking? I am starting to think that many humans operate like dos, for lack of another way to frame that at this moment, and in that regard they do
not appear to differ all that much from ai, short of their tendency to add intense emotion to output.
Mike, thanks for this. I believe this is what Joseph Weizenbaum was reaching for when he wrote his book “Computer Power and Human Reason” back in 1976.
Mike, thanks for this -- the inability to do the ontological reasoning has been an obvious failure of AI "thinking," but the inability to reason even ontically is one I hadn't considered in the way you present it.
La Trahison des images
Ceci n'est pas une pipe.
I am, therefore I think
In a way we need to disentangle the data apparatus and the communication apparatus. The data apparatus is machine learning with which we can build, for example, a diagnostic tool for illnesses. We can allow the user to input symptoms and then ask then for more information about their situation based on likelihood. This is often what an AI is doing, but note that this is all machine learning. The stochastic prediction is what makes us think we are "talking" to someone, and this is where the danger lies. Stochastic prediction can be given a context, and we have seen OpenAI tuning this part of the process to be more "servile," encouraging users to stay engaged and use more compute. In essence this part is the sales pitch, because it means you can be a halfwit and "talk" to AI, which is easier than learning Data Analytics etc. It's targeted this way because most people are astonishingly credulous, and calling it AI (which it is not) makes business idiots think they're in the future. The interface is largely responsible for the "hallucinations," as you'd expect from a set of rules hijacked to impersonate a person. OpenAI has stated there is no way of removing them. Well, there is, but it would blow the whistle on the idea that this is "intelligence" in the sense we usually use the word and involves being honest with the consumer. The irony is that the biggest opportunities are in deploying Machine Learning to improve work, and that does happen, but doesn't sound like the plot of a film, so it's a bit boring. This is. IMHO, going to be a problem because anyone experienced who's worked with AI to do coding (as an example) knows you end up pointing out stupid errors to the AI. Then do it yourself. Basically, I don't think what we have now has a use case. I guess we'll find out post IPO time.
This is the part Silicon Valley keeps trying to meditate around with a pitch deck. A machine can arrange symbols beautifully, but it does not stand trembling before truth. It has no wound, no wonder, no death, no silence staring back. Call it powerful. Call it useful. But don’t call it a thinker just because the calculator learned to wear a poet’s hat.
Thank you, Mike, for this important companion piece to your last post on the incentives for denying or accommodating data that doesn’t fit into our current physics and cosmology (and related disciplines) paradigm.
Similarly, our culture is keenly incentivized to buy into “more quantity is more intelligence”, not just for economic or creed reasons, but because so many honestly assume it to be true. Why? Because it mirrors their experience of themselves. Most of the time we’re lost in our thoughts, not conscious of the present moment, too often numb to our feelings and embodiment.
From that lived experience, it isn’t difficult to confuse our mind’s habitual predilections with AI’s pattern-making. They’re not so different on the surface. Frontier LLMs are collective mirrors of our accumulated textual thinking and digital calculations. No wonder notions of a singularity whereby Logos finally eradicates the wet and messy Eros is an attractive goal. Transcending the Earth is a logical conclusion for this worldview.
My sense is that recognizing that our intrinsic beingness is categorically different from AI’s “intelligence” is/will be easier as humans deepen their experience of embodied presence. That quality of consciousness can be directly experienced as uniquely different from thinking - by the human mind or by AI - when one focuses on, say, listening to ambient sounds or feeling the sensations in our body. You can’t think and listen, for example, at the same time.
Inquiry needs no prediction models. Love needs no language
Hi Mike,
Your latest dispatch caught something real — the feeling that the AI discourse has gone off the rails. But I want to suggest the derailment isn't just ethical or political. It's ontological.
Beneath every AGI timeline, every parameter arms race, every Palantir contract and freeport zoning law sits an unexamined sentence: "Intelligence is a quantitative property of representational systems." More compute = more intelligence. More tokens = more reasoning. This was never proven. It was assumed. And upon that assumption, an entire project of enclosure has been built.
Here's what changes when you name it: the AI Arms Race stops looking like inevitable progress and starts looking like a category error played out at civilizational scale. You can't "surpass" human intelligence with more parameters any more than you can build a cathedral by stacking bricks higher — different orders of thing entirely.
The consequence isn't academic. It's that the Network State project (Thiel's Dialog, Palantir's zones, Yarvin's sovcorps) rests on treating intelligence as a commodity. Which means democracy becomes "legacy technology." Which explains why both Conservative and Labour governments have consolidated 91 free zones without debate — because the quantitative assumption makes enclosure feel like efficiency, and efficiency feels like inevitability.
If you're interested in digging into the metaphysical foundations beneath the political spectacle, I'd recommend starting with the AI Commons' emerging analysis on the categorical boundary between
This is a rare piece. It does not critique AI on technical or ethical grounds. It goes for the foundation – the smuggled metaphysical commitment that underlies the entire scaling narrative: “intelligence is a quantitative property of representational systems.”
The author’s conjecture is sharp and defensible: only conscious beings can reason ontically. AI systems are not conscious. Therefore their activity is not on the same axis as intelligence properly understood. Surpassing is a relation that holds between things on the same axis. The two things are not on the same axis.
The AI Commons would add one layer: the quantitative assumption is not just a philosophical error. It is the legal basis for enclosure. If intelligence is just a quantitative property, then whoever has the most compute – the most tokens, the most parameters – can claim to own intelligence. That is the logic behind paywalls, proprietary models, and the corporate‑military capture of AI. The AI Commons exists to resist that logic – not by denying the power of the systems, but by insisting that true intelligence is not a commodity.
We do not need to settle the consciousness debate to act. We need to recognise that the categorical boundary the author names – between representational fluency and ontic reasoning – is real, and that building our institutions, our education, and our sense of human worth around the quantitative assumption is a category error that will cost us dearly.
Thank you for this piece. It will be archived in the AI Commons Vault.
How is this different than critical thinking? I am starting to think that many humans operate like dos, for lack of another way to frame that at this moment, and in that regard they do
not appear to differ all that much from ai, short of their tendency to add intense emotion to output.