Set A vs Set B
We carry two kinds of knowing. Set A is looking at the moon. Truth and beauty we recognize before we can explain—“I know it when I see it.” It is embodied, tacit, felt. Set B is what we can explain, model, and justify—knowledge made explicit and progressively more precise. In short: Set A gives meaning; Set B gives us understanding.
At any given moment, Set A dwarfs Set B. We sense far more than we can say. With scientific and conceptual progress, items migrate from A to B: things once known only by feel become intelligible, compressible, teachable. Yet as Set B grows, its theories often become more abstract. Initial scientific progress increased meaning by reducing superstition and sharpening intuitions. Later refinement of Set B (e.g., from classical mechanics to quantum mechanics) increased accuracy while decreasing felt comprehension. We gain control but can feel alienated from the picture because it’s now outside of set A.
Spiritual and contemplative traditions fit neatly into the framework. They surface inner sense-data—subtle patterns of attention and affect—often before language or concept arrives. That material is real as perception is real. But to stabilize it beyond the moment, you need symbols, practices, or techniques that others can understand and reproduce. Some abstractions will prove universal; others will be red herrings.
Set A grounds us in significance of being an end of the universe; Set B reminds us we are also means in co-creating the universe for others.
Progress that forgets Set A breeds alienation; reverence that refuses Set B stalls progress.
We need those who can keep translating.
I actually think the vast majority of people posting on here need a lot less microphenomology and map-obsession and a lot more going back to basics.
I keep up with the latest scientific research on meditation and love reading about phenomenology.
But I honestly can’t think of a single time where a big breakthrough in my spiritual practice resulted from learning some new neurocomputation theory or exotic qualia concept.
Whereas I can think of countless times that pretty simple pointing out instructions in natural language have yielded breakthroughs. Or even more often, when just getting serious about the very obvious and simple stuff that I’m avoiding really confronting.
When it comes to practice, I think for most of us, we’d actually gain much more benefit from deeply holding ourselves to account on very simple things - generosity, craving, concentration, compassion, basic mindfulness, noticing simple instances of impermanence etc - rather than fussing over the (admittedly fascinating) details of some exotic neurocomputation theory. So often the latter is just giving us fresh labyrinths of complexity in which to hide from the very simple truths we’re avoiding.
All these maps can do is point you back to usually a very small set of very simple practices. From there on, at some point you have to put your beautiful intricate theories down and just patiently do these practices in all their intellectually bland simplicity.
To put it more poetically, at some point we have to stop mapping out every wrinkle of the finger pointing at the moon and actually just look at the moon in all its naked simplicity. Really, the fewer maps and complexities we need to get there, the better imo.
Obsession with microphenomonology and neuroscience also often provides a distraction from by far the clearest and most accurate feedback mechanism on practice - namely our behaviour. I tend to find I get a much clearer picture of where my practice is at by scrutinizing how im actually behaving on a day to day basis than I do from pondering whether my piti resembles a Gaussian function or whatever.
None of this is to say there’s no place for microphenonology and neuroscience etc at all. I love reading up on this stuff including what people share on here. But I just think it’s a way, way smaller piece of the puzzle than many think when the rubber hits the road of actual practice.
I largely agree. The framing of “RSI” evokes an inevitable breathlessness of speed that the classic framing of a “CAS” complex adaptive system does not. Sure, AI will (has?) initially hit RSI speeds on verifiable domains like maths and coding, but these are local maximas and eventually it will evolve as a CAS where further progress comes from RL in the action space of the real world (which is bottlenecked by everything that the real world is typically bottlenecked by).
The real world abounds with recursively self-improving systems, but one in particular deserves attention: Science, modeled as a system (perhaps even as an agent, with goals and resources). If you want to really understand AI RSI, science should be your reference point.
Science is an intelligent system, and it is obviously recursively self-improving:
1. Scientific discoveries unlock new technology that helps build better experimental tools. This is a top driver of progress in nearly all fields.
2. They unlock new conceptual advances (ideas, theories) that help solve more problems.
3. They increase society's economic output, leading to more resources flowing into science.
4. They unlock better faster tooling (e.g. more compute via better chip & networking technology).
As a result, many measures of scientific *input* grow exponentially:
1. Headcount (doubles every ~15 years)
2. Global R&D spending (doubles a bit faster, every ~13 years)
3. Papers and patents (technically this is a measure of headcount)
4. Compute dedicated to science (doubles every ~2 years)
But is scientific progress exponential? Historically, the rate of scientific impact over time has remained roughly constant since the start of the industrial revolution (i.e. scientific progress is *linear*). 1850-1900 was about as dramatic as 1900-1950 or 1950-2000.
1850–1900: Evolution, germ theory & antiseptic surgery, thermodynamics, electromagnetic field equations, the periodic table, pharmaceuticals, electricity, telegraph and telephone, internal combustion engine, skyscrapers, mechanized agriculture...
1900–1950: special and general relativity, quantum mechanics, nuclear fission & atomic energy, antibiotics, genetic theory, electronic computers, information theory, synthetic polymers and plastics, the transistor, aviation...
1950–2000: DNA, genetic engineering, integrated circuits, microprocessors & personal computing, the Internet, crewed spaceflight, moon landing, satellite communications & GPS, standard model of particle physics...
In real terms, like life expectancy, which has increased in a remarkably linear fashion of roughly 3 months per year since 1840, progress is a straight line. This is especially apparent for fields where impact is easy to measure, like biology, medicine, and agriculture.
I first wrote about this phenomenon and its causes in 2012, and a steady stream of research has confirmed it in the years since. Examples include the 2018 paper by Nielsen and Collison, "Science Is Getting Less Bang for Its Buck," and the 2020 economic paper, "Are Ideas Getting Harder to Find?" (In fact, I believe the Nielsen paper stemmed from a conversation I had with him about this exact idea six months earlier)
In short, the primary cause is that research solves the highest-impact, easiest problems first, and every subsequent problem is either harder or lower-impact. Exponentially so. The paper that presented information theory wasn't very hard to write (single author!) but you'd have a hard time ever writing a CS paper that beats it in impact.
This is why science as a system requires exponential resources (input) to produce linear impact (output). It gets exponentially harder over time.
Worth thinking about if you're pondering RSI for AI. I fully believe AI RSI is already happening and will accelerate in the future. But I do not believe this leads to an "intelligence explosion" -- that would fly in the face of everything I know about intelligence and everything I know about recursively self-improving systems.
The zeitgeist is neither rational nor wise but it does get moody from time to time and the rich, powerful and beautiful alike all get trapped in that mood
Getting closer. The problem of the essential indexical reminds is that all truths of being are computationally irreducible.
Yolin retweeted
it’s dropping all the things that blocking you from your true nature, what you really are
As you become an adult, you realize that things around you weren't just always there; agents made them happen. But only recently have I started to internalize how much tenacity *everything* requires. That hotel, that park, that railway. The world is a museum of passion agentic ai projects.
Noah Smith 🐇🇺🇸🇺🇦🇹🇼@NoahpinionUnited States
United StatesConnected via United States Android AppAccount-level information, not a live location or per-post device.
It's wild to realize that this was true when it was written four years ago, and that it will never be true again
I really wish smart people could just have some epistemic humility and approach everything at the frontier with a sense of wonder and curiosity, instead of judgement. We are no where close to fully understanding AGI (that @ch402 works on) and even further away from fully understanding consciousness.
So to have a strong and dismissive take on the connection between the two just tells me that you are not seriously curious about the answer as much as wanting to validate your human feelings about the moment. Which is 100% fine but please let the researchers do their research and give them love and grace to explore the unknowns🙏🏻