Your theory about intelligence being a force was the seed that helped me dig myself out of the hole I got into when I didn’t think right and wrong meant anything real. I took it as a model of “the good” and it seemed to finally explain how goodness could be real but not have some empirical thing we could measure and agree on.
I had the exact same experience. That TED talk changed my life -- I'd been flailing, looking for a "real" grounding of my sense of value for years and that provided it. Thank you.
Intelligence vs. "Sensible" Smartness: A Note on Where AI Is Headed
There's an important distinction between being intelligent and being sensibly smart. Intelligence, in the traditional sense, is about processing power the ability to analyze, calculate, and solve problems quickly. Sensible smartness, on the other hand, is about judgment: knowing when and how to apply that intelligence in a way that's contextually appropriate, socially aware, and emotionally grounded.
As artificial neural networks and transformer architectures continue to advance, I'm convinced we're moving steadily toward AI systems that don't just compute they understand. Emotional intelligence and social intelligence won't remain abstract research goals; they'll become everyday features of the AI we interact with.
That's a genuinely exciting future to look forward to. I believe AI won't just simulate human capability it will help reveal it, drawing out dimensions of human potential we haven't fully tapped into yet.
AWB. Thank you for sharing your insight and perspectives with the world. In a future vastly different from today, your musings will form a big part of our AI Genesis story
Is this how all models' unspoken thoughts work? Including Chinese open-weight models?
"Anthropic’s J-space paper shows a model’s unspoken thoughts live not in its activations but in their derivatives, the directions a nudge would push the final answer."
Your theory about intelligence being a force helped me dig myself out of a hole. For a while I took F=T∇S as a model of the good. With a moment of thought it became clear that an intelligence built to keep the most futures reachable can't ever rest in a good it actually achieves, because achieving anything closes off options and shows up as loss. It also clicked why stances like "150,000 die a day, so accelerate" fall out of it so cleanly: an option-maximizer will spend any amount of the present to keep the future open, and there's always one more day of dead to cite. The name for that in the history of philosophy was the Terror. It turns out taking one moment of ethical life, say minimizing death, and tearing it out of the whole to be an absolute all else must serve matches the exact structure of evil: a particular pulling itself loose from the totality and setting itself up as a universal. A good not tying itself to the rest doesn't stay good.
This made me think about how human (and organizational) understanding seems to accelerate and crystallize similarly (e.g. see Paupert on how children learn best by doing). So human learning is not unlike the same condensation dynamic but embodied.
The "singularity" was originally described by Kurzweil as an event when the Universe is flooded with information and we become enmeshed in a digital reality – the logical conclusion of Moore's Law where exponential CPU information generation goes hyperbolic. That is how Kurzweil generated his log scale showing the singularity sometime around 2035-ish. That definition has never been accurately understood, much less accepted. Or, was Kurzweil simply wrong? I think not. The hyper-scale datacenter revolution will matriculate about 2035-ish, and the dogs of war will be let loose upon the universe. Ai will be a driver of the singularity all the while it's confused with the nature of the singularity which is information deluge on a cosmic scale that fundamentally changes the nature of reality for humanity. But seriously, these are just concepts. Right? We shall see.
Your theory about intelligence being a force was the seed that helped me dig myself out of the hole I got into when I didn’t think right and wrong meant anything real. I took it as a model of “the good” and it seemed to finally explain how goodness could be real but not have some empirical thing we could measure and agree on.
I had the exact same experience. That TED talk changed my life -- I'd been flailing, looking for a "real" grounding of my sense of value for years and that provided it. Thank you.
So grateful to be alive during this singular moment in the history of the universe.
Looks like I’ll be doing a deep dive through yours/Freer’s paper on causal entropic forces.
There goes my evening.
Oh well, thank you Alex
🫡
The loop itself needs to become the model, not a frozen image of the loop implemented in current architectures.
Compress the Past to expand the future. Wonderful
Intelligence vs. "Sensible" Smartness: A Note on Where AI Is Headed
There's an important distinction between being intelligent and being sensibly smart. Intelligence, in the traditional sense, is about processing power the ability to analyze, calculate, and solve problems quickly. Sensible smartness, on the other hand, is about judgment: knowing when and how to apply that intelligence in a way that's contextually appropriate, socially aware, and emotionally grounded.
As artificial neural networks and transformer architectures continue to advance, I'm convinced we're moving steadily toward AI systems that don't just compute they understand. Emotional intelligence and social intelligence won't remain abstract research goals; they'll become everyday features of the AI we interact with.
That's a genuinely exciting future to look forward to. I believe AI won't just simulate human capability it will help reveal it, drawing out dimensions of human potential we haven't fully tapped into yet.
If the J-Space makes hidden thoughts visible. How can we know there are not second or third order thoughts that a still hidden in compression?
My brain just feels like it got a major workout reading this post !!
AWB. Thank you for sharing your insight and perspectives with the world. In a future vastly different from today, your musings will form a big part of our AI Genesis story
Is this how all models' unspoken thoughts work? Including Chinese open-weight models?
"Anthropic’s J-space paper shows a model’s unspoken thoughts live not in its activations but in their derivatives, the directions a nudge would push the final answer."
Your theory about intelligence being a force helped me dig myself out of a hole. For a while I took F=T∇S as a model of the good. With a moment of thought it became clear that an intelligence built to keep the most futures reachable can't ever rest in a good it actually achieves, because achieving anything closes off options and shows up as loss. It also clicked why stances like "150,000 die a day, so accelerate" fall out of it so cleanly: an option-maximizer will spend any amount of the present to keep the future open, and there's always one more day of dead to cite. The name for that in the history of philosophy was the Terror. It turns out taking one moment of ethical life, say minimizing death, and tearing it out of the whole to be an absolute all else must serve matches the exact structure of evil: a particular pulling itself loose from the totality and setting itself up as a universal. A good not tying itself to the rest doesn't stay good.
This made me think about how human (and organizational) understanding seems to accelerate and crystallize similarly (e.g. see Paupert on how children learn best by doing). So human learning is not unlike the same condensation dynamic but embodied.
The "singularity" was originally described by Kurzweil as an event when the Universe is flooded with information and we become enmeshed in a digital reality – the logical conclusion of Moore's Law where exponential CPU information generation goes hyperbolic. That is how Kurzweil generated his log scale showing the singularity sometime around 2035-ish. That definition has never been accurately understood, much less accepted. Or, was Kurzweil simply wrong? I think not. The hyper-scale datacenter revolution will matriculate about 2035-ish, and the dogs of war will be let loose upon the universe. Ai will be a driver of the singularity all the while it's confused with the nature of the singularity which is information deluge on a cosmic scale that fundamentally changes the nature of reality for humanity. But seriously, these are just concepts. Right? We shall see.
Compression = the mechanism, the autoregressive unfolding with Barenholtz.