sixo

Psychological state is key, and I think you're understating it.

The human mind is always passively working on problems. We are not necessarily aware of how much working memory is "tied up" on background computations, but when it's a lot, we will feel a natural aversion to new learning, it will be exhausting, and we will lack for curiosity. This is fine, working as intended.

When in this state, what one usually needs to do is a different part of the learning process: not absorbing new information or abstractions, but making sense of what one already knows, working out contradictions, assembling larger-scale theories--synthesis, basically. "Learning" is only one part of the arc of "thinking".

Usually this is best done by writing, and usually comes easiest when trying to convince someone else or what we believe. If you can't find the energy to learn, try writing instead.

That's a nice argument, though it doesn't hold relativistically

Kinetic energy is, strangely, quite a bit like a least squares cost function in an optimization problem. The "dt"s in "dx/dt" hardly matter; it basically represents "dx^2" between the current state and the next.

I read this kind of essay as a certain part of the arc by which new thoughts are formed: an act of large-scale pattern matching, laying out a bunch of cases which resemble each other, searching for the essential basis of the resemblance.

To post such a pattern allows the thought process to become distributed. Perhaps someone else will see the insight.

things which help other people make money are overvalued relative tothings which are valued intrisically

I suspect that, in reality, it is the indignity of poverty which motivates people to take up arms against each other. So long as dignity is retained, poverty may be emotionally bearable (perhaps to the point of actual starvation, when dignity becomes unsustainable).

To hold this view you have to think of information as "not real", not like "real" molecules and receptors, the mind as distinct from the body, and then restrict the legal definition of harm to only "real" things.

This is an odd thing to do, because :

- information is real, it exists in the universe.

- the harm of social media is real, as measured by many of the same measures as the harm of smoking

Why not do something about ads? No, that's a good thought, we should do that too.

I think a decent conceptualization here is "psychic damage", as in a video game. These things deal a lot of it.

This ought to change your mind about Ruby!

Feels fairly likely, though I'm skeptical the Jeeves name recognition is much besides a novelty now.

Missed opportunity to name an LLM "Jeeves" and finally live up to the vision.

The right read here is to realize that psychology alone is not the basis for moral concern towards other humans, and that human psychology is, to a great degree the product of the failure modes of our cognitive machinery, rather than being moral.

I find this line of thinking to lead to the conclusion that the moral status of humans derives from our bodies, and in particular from our bodies mirroring others' emotions and pains. Other people suffering is wrong because I empathically can feel it too.

We as a society accept the insurance system as an implementation of "funding healthcare" because market capitalism is supposed to lead to lower prices, fair allocation of scarce resources, and innovation, among other things. That is, the insurance industry is a market solution to a moral problem.

If insurance companies then can wiggle out of covering pre-existing conditions, they're no longer solving the moral problem they were brought into the world to solve, and now we need some other solution to solve the rest of it. Then, whatever that other solution is, it's solving the hard part, so why not extend it to solve the whole thing and cut the insurance middlemen out of the economy entirely? What are they even doing at that point besides extracting a rent?

(This is one answer among many good ones to what is really a bad-faith question—health-insurance is not a lot like fire-insurance at all)

HN people always try to do this cute rhetorical gesture where you take a thing and say "hmm nice idea what if we called it <thing that already exists>", but they like this joke so much they get baited into doing it in dumb ways like this one.

A coworking space in every building != a WeWork. There's a big difference between these! You could implement the former by opening a million WeWorks but that doesn't sound good at all; residential apartment buildings already have common areas, free to residents, they would simply have to be reimagined slightly.

I don't see how that's necessary at all. All the arguments that WFH might be a good idea in the first place would still hold.

I would love to have a coworking-space-on-every-block (or in every building) where all the WFHers can go to be around other people (just not the coworkers)

A reputation moat is still a moat. It seems to me that Lego prices will drop as soon as they are forced to by competition, and not before, and this is fine.

Prices are constrained by demand moreso than by cost of production. Lego pieces are expensive because they can be, they still sell, and this is largely due to the quality. As long as the quality moat persists, they can charge as much as people will pay, and--good for them!

That you personally would prefer lower prices does not mean they "should" be lower. Those lower costs of production, to Lego company, "should" mean higher profits, not lower prices, and again--good for them!

It appears that the real lesson here was to lean quite a bit more on theory than a programmer's usual roll-your-own heuristic would suggest.

A fantastic amount of collective human thought has been dedicated to function approximations in the last century; Taylor methods are over 200 years old and unlikely to come close to state-of-the-art.

There's also a trip to Hawaii in it for you.

Your dismissal of moral concerns is not convincing.

Imagine a world where the only energy you do is use was generated by a stationary bike you had to ride yourself. You would, generally speaking, use that energy differently than energy you would pay for--you would generally reserve your effort for worthwhile things, and would be averse to farming energy yourself just to power frivolity or vice. How you determine what to put your energy into would explicitly be a moral question.

Instead in our world we an abstractions conceals the source of the energy. But if the moral concerns from the first world had any weight, they haven't lost it now; if energy is anything short of completely free we should by the same logic be averse to expending energy on worthless work or vice. The human being is not a utility monster, but something very different, and moral questions of this sort are central to how it navigates the world, they should not be dismissed.

Not every conversation is about AI. This one isn't.

this is false, it's been around for a while

I think the issue with "modeling" is really a human one, not a mathematical one.

It's helpful sometimes to think of our collective body of mathematical knowledge as like a "codebase", and our notations and concepts as the "interface" to the abstractions at play within. Any software engineer would immediately acknowledge that some interfaces are FAR better than others.

The complex numbers numbers are one interface to the thing they model, and as you say, in a certain sense, it may be the case that the thing is C. But other interfaces exist: 2x2 antisymmetric traceless matrices, or a certain bivector in the geometric-algebra sense.

Different interfaces: a) suggest different extensions, b) interface with other abstractions more or less naturally, c) lend themselves to different physical interpretations d) compress the "real" information of the abstraction to different degrees.

An example of (a): when we first learn about electric and magnetic fields we treat them both as the "same kind of thing"—vector fields—only to later find they are not (B is better thought of as bivector field, or better still, both are certain components of dA). The first hint is their different properties under reflections and rotations. "E and B are both vector fields" is certainly an abstraction you CAN use, but it is poorly-matched to the underlying abstraction and winds up with a bunch of extra epicycles.

Of (d): you could of course write all of quantum mechanics with `i` replaced by a 2x2 rotation matrix. (This might be "matrix mechanics", I'm not sure?) This gives you many more d.o.f. than you need, and a SWE-minded person would come in and say: ah, see, you should make invalid states unrepresentable. Here, use this: `i = (0 -1; 1 0)`. An improvement!

Of (b): the Pauli matrices, used for spin-1/2 two-state systems, represent the quaternions. Yet here we don't limit ourselves to `{1, i, j, k}`; we prefer a 2-state representation—why? Because (IIRC) the 2 states emerge intuitively from the physical problems which lead to 2-state systems; because the 2 states mix in other reference frames; things like that (I can't really remember). Who's to say something similar doesn't happen with the 2 states of the phase `i`, but that it's obscured by our abstraction? (Probably it isn't, but, prove it!)

I have not given it much more thought than this, but, I find that this line of thinking places the "discontent with the complex numbers in physics" a number of people in this thread attest to in a productive light. That dissatisfaction is with the interface of the abstraction: why? Where was the friction? In what way does it feel unnecessarily mystifying, or unparsimonious?

Of course, the hope is that something physical is obscured by the abstraction: that we learn something new by viewing the problem in another frame, and might realize, say, that the interface we supposed to be universally applicable actually ceases to work in some interesting case, and turns out to explain something new.

Even in the US we see cities becoming desirable place to live when they successfully cultivate a film scene, or an art school, and being dead when they don't. But this feels like a better approach than a basic income (which is an invitation to idleness)--make it easy to use the environs for film, streamline permitting, provide cheap capital, solicit locals for public installations.

The point is not to make better predictions of the things we already know how to predict. The point is to determine what abstractions link the things we don't presently understand--because these abstraction tend to open many new doors in other directions. This has been the story of physics over and over: relativity, quantum theory, etc, not only answered the questions they were designed to answer but opened thousands of new doors in other directions.

Here's one: why does sunlight appear "sharper" or "harsher" during cold winters?

The reason, I believe, is that cold air tends to be drier, and drier air scatters less, leading to more of the overall flux you receive coming straight from the sun rather than from the diffuse source of the sky around you. But I'm not certain of this.

It might be "never-ending", but you're going to encounter edge cases in approximate proportion to the rate at which they actually occur. Anyway, the hope would be to learn behaviors which generalize, not to respond to each edge case ad-hoc; the edge cases provide out-of-sample tests of generalizability.

Ideally it would learn a relationship between the sensor input and the correct actions, even if the sensor input is not realistic for the GG in reality.

Well it certainly helps,doesn't it? This system is going to encounter more edge cases than a single human ever would. Hopefully the lessons from known unknowns generqlise to unknowns. And once they've been seen once they took can become part of the corpus.

Certainly commensurate to the price. It's up to the companies to bring the cost under the price.

AFAICT, fears of the marginal costs of LLM inference being high are dramatically overblown. All the "water" concerns are outlandish, for one—a day of moderately heavy LLM usage consumes on the order of one glass of water, compared to a baseline consumption of 1000 glasses/day for a modern human. And the water usage of a data center is approximately the same as agriculture per acre.