darkmighty

Yea, but it still doesn't make sense. I mean, imagine Stephen Hawking really became incommunicable. He would still be very much conscious, without exhibiting any behavior. Even disposition of behavior doesn't make sense, because behavior is simply impossible. We're back to talking about the internal state of the brain/mind, and not simply behavior disposition.

So the idea of behavior disposition being central to consciousness is clearly silly. It is clearly a complicated function of the workings of the entire brain as a system -- even though the sensory and actuator (i.e. behavioral) systems are important to the individual, they are by no means exclusive or central.

Could you perhaps summarize Dennett's counter argument to Chinese Room?

I think a more realistic (and worrisome) scenario is war brekaing out in South East Asia. Fabs should be clear economic targets for opposing forces. There are a number of conflicts which would likely lead to severe disruption on global semiconductor market.

Oh yea my bad, I meant the set of all halting functions (which is (program->halts)) -- those are incompressible (again actually finitely compressible). The set of all programs can be trivially enumerated depending on the Turing machine (it's usually just a form of the set of binary natural numbers).

Note I did not prove incompressibility, only that there exist some incompressible functions; indeed there are as many functions as there are real numbers (in cardinality), so almost every function is incompressible! (imagine all different strings of random bits) For the halting function I believe the proof is a little involved and can be found on the book "Elements of Information Theory".

I find wikipedia's proof (https://en.wikipedia.org/wiki/Halting_problem#Proof_concept) not quite illuminating though, since it is a basic proof by contradiction. Those tend to leave you wondering if you cannot replace the halting function halt(x) with a self-reference-free function halt_noself(x). Compressibility give a wider intuition that sure, there are special subsets of the halting function (program->halts) that are compressible, but in general it will be incompressible, and therefore a program will only be able to decide finitely many sequential programs -- and this table can only be done by a "luck-based" search: you test the programs to see if they halt. The ones that do not halt in any reasonable amount of time, you may try searching the space of proofs that it doesn't halt. This search is again undecidable, because at some point there will be a program which doesn't halt but that you cannot prove.

So in some ways the best you can do is grow the list with holes such as "program 1 halts, program 2 does not halt (proof), ...., program k-1 halts, program k ???, program k+1 halts, ..." -- you may never be able to fill any particular hole. You can also build a set of heuristics to accelerate both the simulation of the programs (to show they halt) and the evaluation of proofs (to show they do not halt).

This is essentially the process by which we do mathematics, or even computer science :) We use a growing set of heuristics to prove statements we deem "interesting" or "useful", and sometimes simple statements will come up (such as Collatz conjecture, Goldbach's, Twin primes, Riemman hypothesis, ...) that we cannot prove and we do not know if we will ever be able to.

This is a valid method of deciding whether a program hangs. Another very simple approach is searching for global state recurrences (if your program goes back to exact same state at least once, and it is deterministic, then it never halts).

Note this has no bearing on the Halting problem though (famously proven impossible). You cannot decide whether an arbitrary very complicated program halts -- it may do all sorts of weird stuff that doesn't involve any of the previous behavior (having necessary subfunctions never terminate or global recurrence) -- it could try counting up to infinity until a certain complicated binary function of the number is true, then you have to decide the truthiness of this arbitrary function. In general you face a self-referential problem: imagine you try deciding whether a program of size N halts if it runs for too long. You'd need a program (say of size K) that produces a function f(N) of maximum running time for programs of size up to N -- then you could just check whether a program runs for more than f(N) steps. But by definition your program cannot run for more than f(K) steps, thus it cannot output a number greater than 2^(f(K)). Thus your program can only decide finitely many instances, and is not an algorithm (instead it is a sort of compressed look up table).

The proof of the Halting problem undecidability is more general, but uses similar self-referential arguments.

I like to think of it this way: an algorithm is really a compressor function. It compressed an infinite set of (input->output) relations into a finite program. But some infinite sets are inherently incompressible -- think a string of random numbers -- and in particular the set of all programs. Incompressible (actually finitely compressible) sets must exist because otherwise you would have infinitely many different sets that are output of finitely many programs, which is obviously impossible (this is the pigeonhole principle).

I believe c/4 is indeed the fastest possible diagonal speed for a spaceship, does anyone know if a proof exists?

Definitely. Ozone is a significant irritant (although it also has an "air freshening effect", see air ionizers), to produce significant airflow I believe you'd need very large electrodes, which would produce lots of ozone. The electrodes need to be large because the voltage is limited by air's breakdown voltage, so you can't just pump up the voltage like you can (for the most part) just increase a fan's velocity.

For lack of friction/bearings a magnetic bearing fan seems like a much better option. The strength of this technology is the great exhaust velocity, which is great for space applications.

Would it be a good idea to have major kernel forks targeting different sizes? Say for example {kernel_tiny, kernel_small, kernel} targeting {<4MB,<100MB,<inf} respectively. Although as you mention I'm not sure it is really a wise choice to use Linux for really small and simple systems.

A few points: data movement is expensive, but the reasons are not so simple:

1) x68s RAM needs to move huge amounts of data cheaply and extremely reliability. This requires a lot of power to fight noise. Once you're allowed to have a few bit flips, the bus voltages can be made much power directly lowering losses.

2) A significant portion of the large energy costs is actually in computing associated with memory movement! (according to Sohmers from REX Computing, about 40%). Encoding, queuing on processor, Decoding, fetching, queuing on the memory controller, and caching again on the CPU takes a lot of logic.

I think it's quite possible this approach is superior for low precision floating point applications that currently use CPUs/GPUs. Note this allows lowering voltages for the whole system, including interfaces and memory.

Also, data movement doesn't dominate the energy cost of all applications; there are plenty of operations still compute-bound (e.g. some evolutionary methods), it's just that the applications more in vogue are like that. There are plenty of applications that are currently done in DSPs and ASICs because of heavy computational costs (error correction decoding as mentioned in the article) that might benefit from this.

---

As far as I know, it is still an open question whether it is possible to generalize this approach for data-efficiency. I have a few ideas I've been playing around with but none that seem to work yet.

I read a while ago a researcher's website specialized in puzzles. He claimed (informally) the only really "deep" puzzles that can keep you hooked are based on NP-hard (or harder) problems. Polynomial-time puzzles according to him just depend on finding a "trick" after which you can solve most puzzles relatively quick. Specially O(n) puzzles. Which is the case for this planarity puzzle! I did enjoy the few ones I did, perhaps I didn't "get" the linear-time algorithm right away.

NP-hard puzzles otoh, are hopeless for hard instances and general algorithms. Both the designer has to give it more thought on making instances that are not worst-case, and you have to rely not on general-case algorithms but an increasing set of heuristics and intuition to solve. Brute force is completely hopeless.

The same goes for two-player games. If the game can be solved in polynomial time, one of the players will find a trick (the polynomial time algorithm essentially) to always win (or depending on the game always draw).

I find this [type of] lock quite usable, and much more convenient thne keys:

https://www.amazon.com/Schlage-FE575-CAM-619-ACC/dp/B001RFDB...

Digital so I don't need keys, but still keeping it as simple as possible. Most have a power backup so you can power it from a 9V battery externally. I still wish it wasn't necessary to change batteries, but it's not too inconvenient given they last a couple years, timewise well worth not dealing with keys.

Like Newton's Principia (from 1684) looks like Norse mythology? It is still a valid model except for a few minor errors, and almost halfway to 1000 years. Better yet might be Greek mathematics, e.g. Euclid's Elements which is more than 2300 years old now. The language is different of course, but the content itself reads a lot like a good proof-oriented geometry textbook.

I love this kind of argument. It is of course very simple that every real number is uncomputable -- real numbers have greater cardinality than the integers. So if you had a program (which can be described by a finite sequence of integers) capable of producing each real, you would have a finite integer sequence for each real, contradiction.

Note though that quantum mechanics adjusts this notion of real numbers in classical mechanics. A full quantum description of our world in theory requires only finitely many bits, because the quantum state is finite. But then measurements in quantum mechanics are non-deterministic, so impossibility of perfect prediction is still true :P

> It is a zero-sum game, and increasing our consumption will disrupt other processes.

You need to define what you mean by "being a zero-sum game". For example, in terms of entropy (which is a possible metric in the context of "dissipation") it isn't a zero-sum game -- certain processes increase entropy more, some less. In terms of total instantaneous thermal absorption, which defines the temperature of the Earth, it also isn't a zero-sum game -- If there was a white surface and it is made black, then the temperature of the Earth will rise; the same goes for emitting CO2 which in effect increases the global absorptivity (energy that wouldn't be absorbed if the Earth didn't contain as much GHG). Putting solar cells in barren deserts isn't going to disrupt local lifeforms simply because there aren't any. So it seems in almost every definition of zero-sum it indeed isn't a zero-sum game -- which is good!

Note that I didn't claim renewable energy has no environmental cost. It most definitely has. Every activity will have certain impacts, greater or lower, on resource availability, thermal balance, and on other lifeform habitats. The key is pricing adequately those impacts so any damage could be repaired or compensated. The effect of pricing is usually just impact minimization, since the cost of repair or compensation is usually much greater than prevention.

So it is a matter of seeking technologies that give the greatest return (power output in this case) for total cost (environmental and economic) -- and better alternatives than current ones are constantly being discovered, thankfully.

But why would we need to reduce consumption? We need emissions tax -- a tax proportional to the environmental cost. If power consumption carried little or no environmental cost, we'd want to use a huge amount, and that wouldn't be an issue.

There are clock that run on small temperature variations, there seem to be commercial ones with a hefty price tag:

http://www.jaeger-lecoultre.com/us/en/watches/atmos/atmos-cl...

(previous discussion: https://news.ycombinator.com/item?id=16419434)

The problem is those don't have an automatic synchronization mechanism. That's the really tricky/cool part about the 10ky clock.

If you wanted to create, say, a time capsule that'll last for 1000 years with a working watch, I think your best bet would be some ultra-low power microcontroller and an LCD that you could turn on occasionally. Both power and syncronization could be done from temperature variations. Just leave it buried somewhere outdoors exposed to daily (night and day) and seasonal variations, it should be enough to keep time with +/- a few hours at decent probability.

Just make sure to avoid anything that exhibits remotely any degradation, like electrolytic capactitors, and account for erosion of the moving parts.

Some relevant discussion: https://electronics.stackexchange.com/questions/317684/could...

Agreed. It's a disordered, scary place to be in. Not something to glorify and fantasize as a mythical or sensory experience (especially since unlike hallucinogen drugs it's not voluntary, it lasts potentially all your life and you're not in control).

Were you diagnosed?

Do you have neurological issues? (or suffer from risk factors? such as high blood pressure, overweight, diabetes, sedentarity) I wonder if that could be related (because of PKD's eventual tragic death by stroke). As far as my tiny medical knowledge goes schizophrenia hallucinations usually auditory in nature and not very explicit. There's also the question of the effect of the drug supposedly taken during his dentist procedure ( a compound which seems to have brain effects https://en.wikipedia.org/wiki/Sodium_thiopental#Side_effects ).

Of course he was probably in general risk for delusional thinking given his job to constantly devise bizarre plots questioning every aspect of reality. But it does make me curious whether he suffered from a neurological condition, or just a psychiatric disorder such as schizophrenia.

Hope you're doing well, mental disorders are scary :(

Nice website, note by the same source that the US military budget is about $1t per year ($886 billion). Even more insane.

https://www.thebalance.com/u-s-military-budget-components-ch...

For a fraction of that I'm pretty sure we could be well under way of self-sustaining colonization of other Solar System bodies.

If you battery is dead and your engine isn't running, isn't the battery the priority anyway? Having backup batteries to provide initial excitation for generators should not be difficult. They just need to start it, afterwards you can use the output for self-excitation. Worst case, you'd just use a tiny PM generator to bootstrap the process.

I think there are other more significant trade-offs between induction and PM motors. Induction motors/generators are significantly heavier, and have significantly lower torque (i.e. can extract less lower at low speeds when used as generator). Their main attractiveness comes from lower cost, reliability, and decent efficiency at high speeds.

> Indeed, SPECint performance continues to grow exponentially, albeit more slowly since ~2004

Interesting chart! However that looks like moving goalposts. If you allow functions to still be exponentials when the rate repeatedly diminishes, any monotonic function can be an exponential: f(x) = x is an "exponential" with diminishing growth rate log(x). That curve looks approximately logistic to me, as most technologies usually do.

Actually I believe speculative execution theoretically allows increasing linearly single-threaded performance for exponentially more multi-threaded performance -- so a transistor price Moore's law (if not power/density) should allow a continued linear single-threaded growth. The problem is that without power (inverse) scaling, this would also cost exponentially more power. It seems the slight increase in power visible in that chart could account for a fraction of increased single-threaded performance (other factors would be improved efficiency and architectural gains); expect those factors to also stall soon (which would fulfill the "logistic prophecy").

I'm thoroughly impressed by your geometric abilities! I didn't know that, and took me a while to check. Any hints on the puzzle, and as a sidequestion, what tools do you use to figure this kind of question? Just imagination, vector algebra, elementary trigonometry?

Well that's IMO devoiding the word of original meaning then. You're referring to a revolution, which is a well established term, not a bona fide 'Singularity', which comes from mathematics as a point where the speed of change properly diverges -- as would be the case if we had a geometric time series of events with constant improvements.

This usage of the term really originated in the context of rampant intelligence growth (through a supposed explosive self-improvement), see the wikipedia article:

https://en.wikipedia.org/wiki/Technological_singularity

Even AI researchers fall to the Singularity (imo) fallacy, including Schmidhuber:

http://people.idsia.ch/~juergen/history.html

Although at least there's some display of self-skepticism:

"Kurzweil (2005) plots exponential speedups in sequences of historic paradigm shifts identified by various historians, to back up the hypothesis that "the singularity is near." His historians are all contemporary though, presumably being subject to a similar bias. People of past ages might have held quite different views. For example, possibly some historians of the year 1525 felt inclined to predict a convergence of history around 1540, deriving this date from an exponential speedup of recent breakthroughs such as Western bookprint (around 1444), the re-discovery of America (48 years later), the Reformation (again 24 years later - see the pattern?), and other events they deemed important although today they are mostly forgotten."

Which is a little rare, if you know the curious character of Jürgen Schmidhuber :)

Electronics engineering (all engineering as far as possible really) does have a good amount of modularity. So understanding a circuit is about understanding each module, or simply being told their general function. You have a general geometry e.g. a balun or an antenna should follow, but the details very much follow from complex electromagnetic interactions, and I assume are designed through a intuition, simulation and testing loop. So if another engineer needs to slightly modify your design, he will usually just try adding or modifying individual modules to get the desired function -- that could include modifying the geometry of a circuit board analog filter, re-shaping an antenna for different directivity characteristics, etc.

> A key performance metric of computers is their energy dissipation. One contribution to dissipation is friction at the rotary joints in each logic gate. Due to the joint’s small frictional drag, mechanical computers constructed from them can, in principle, dissipate orders of magnitude less power than conventional semiconductor computers, while still operating at relatively high speeds.

This claim seems dubious, perhaps someone has more expertise to comment. The justification

> Operating this lock involves rotation at the joints by up to ∆θ≈1 rad. The model system analyzed in [11] is an excerpt of the links and joints shown in the closeup on the right of Figure 24. From [11, Eq. 2], this rotation dissipates bout 2.4×10−27J per rotary joint when operating at f = 100 MHz.

Seems akin to saying "We operate our microchip at 1 microvolt / 1 pico ampere at f=100MHz, giving 10-26J per operation." -- which seems like a silly aspiration without carefully analyzing noise and quantum mechanical constraints. (without which it would seem almost any computing device could operate at arbitrarily low power).

I think this question is quite complicated, much more than it appears. Many questions come to mind.

Is trivial entertainment completely unjustified? What exactly entails trivial entertainment? What makes acquiring a mechanical skill in a game more or less trivial than reading a novel?

And what exactly does self-improvement constitute? Is it about knowing more facts? Having more skills than our predecessors? Having more developed particular skills? What about ethics and morality, are they amenable for self-improvement or mostly a product of one's experiences? How should ethics and morals be weighted against other areas of self-improvement?

What is the ultimate goal of self-improvement, or is it a goal in itself? If living a long happy life is a goal, should we give into some simple pleasures that make us happy?

Obs: I don't mean to imply that Candy Crush is the pinnacle of human condition, I've never even played it myself to judge. More to emphasize the hazards in analyzing this sort of question.

> such that loss of a single bit means the original key is unrecoverable

That can't be right (I don't know anything about this though). If you control the data (with a non-interactive non-destructive decryption process), a loss of a single bit just means you have to test two possible keys. In terms of brute forcing, the security is the same per number of bits lost. The benefit of using a large key would be that if you lose a certain fraction of your key (say 10%), then that would correspond to more bits (as long as you erase 128 or more bits you'd be fine).

Also beware of key stretching, use it only when absolutely necessary. Key stretching doesn't modify the ratio of work necessary for your encryption/decryption vs the work necessary for brute forcing -- i.e. it doesn't improve the security factor. You're essentially doing an economic defense vs a mathematical one, and hoping that computers won't improve and your attacker isn't willing to spend much relative to what you spent.

You should post this verbatim on social media.

To be fair, the hardcore crypto-anarchists and such would never keep their money on exchanges. Ideally they'd keep money on cold wallets, physically secure.

But also they don't care about usability, which sucks in that case. Being your own bank sucks, pretty much.