HNHacker News
TopNewBestAskShowJobs

WilliamDhalgren

204 karma · joined August 24, 2016

submissionscomments
WilliamDhalgren··on Robots Can’t Dance – Why the singularity is greatly exaggerated (2015)
hm, used to be said for chess too; a uniquely human creative skill.. I love the Go algorithm (there's 3 basic ones in there really), but no aspect of it seems creative.

To try to see it in a grossly simplified step by step fashion: in the abstract, does seem similar to how a human would think of the game though; consider promising moves, where promising is some fuzzy intuition based on the pattern of the board -> for each of those, read how the responses might look like (which the computer does way way more than the human), and what you discover using this suggests other promising moves -> and at some depth of each such read, evaluate the board position; who looks ahead, which too is some fuzzy-ish intuition based on the board patern (though there are various heuristics humans would use when learning this). Computer also evaluates further, a number of fast playouts to the end and mixes this with that pattern-based evaluation.

Is this the kind of thought process you'd associate with creativity? Seems search-based, analytic and strategic; only funny thing is how evaluations are pattern-based nebulous things, but that seems an implementation detail. In some situations, you may come up with evaluation procedures like say material advantage in chess and code the relevant decision procedure, in others, you'd "code" the decision procedure implicitly, by specifying a dataset that exemplifies the desired function, and use statistics to come up with that function. Increases the kinds of decisions you can implement tremendously as you can leverage implicit rather than just explicit knowledge, yes -- but conceptually kinda the same.

WilliamDhalgren··on Robots Can’t Dance – Why the singularity is greatly exaggerated (2015)
heh a decent point, but to be fair the task really isn't "make a picture in the style of <artist>" but "make a picture in the style of this one other picture". I'm not particularly convinced an otherwise naive human could separate style from motif in such a limited setting.

It mainly demonstrates that the network learns rather abstract characteristics of the original picture, to be possible to factor them out. Hardly a way to create new masterpieces, but a nice demo of computer vision advancement.

WilliamDhalgren··on Robots Can’t Dance – Why the singularity is greatly exaggerated (2015)
what? eyes?? really? aren't you confusing the deep dream work with the style transfer work? I remember the dreams thing having eyes everywhere.

Can you show me an example? I'm thinking of something like this: https://github.com/jcjohnson/neural-style though I think there was a webapp to upload your own and try it out that I couldn't find now...

WilliamDhalgren··on New report examines how AI might affect urban life
I completely agree there's a ridiculous amount of work to be done for a human level intelligence -- I'd even speculate that there's little economic need for one too; most economic tasks, even high-level creative and strategic ones are much more narrowly defined, you could make do with systems that simply comply with their master's bidding in a request/response fashion, which seems antithetical to open-ended self-motivated actions of an introspective human being -- I'm not sure about your examples though, as these seem precisely where I've read about significant progress made lately:

- teleoperation is generally orthogonal to creating a robust bipedal gait, and it seems like capture point approaches by BD & whatever Shaft is doing, if one can believe their marketing videos, as well as underactuated systems like DURUS and ATRIAS (and its bird-like successor, can't find the name, C-something) are starting to look quite robust. Some work still to be done, I think but looking quite promising. And this is all just rather classical approaches; nothing machine-learning-y at all. You'd still teleoperate them to navigate and make them do a manipulative task, but not just because it can't walk otherwise; but because those are different domains and problems (like manipulation, autonomous navigation etc). Hell, ATRIAS is completely blind anyhow, all you can do is define foot clearance height and direction of movement.

- deep reinforcement learning based visuomotor control techniques demonstrated last year on the PR2 platform and later on I think also demonstrated on a Baxter robot seem pretty dextrous at such manipulative tasks like folding shirts, and does it realtime, after some trial and error self-training. Granted, didn't see them do exactly that task, but a bunch of other manipulative exercises were demonstrated, realtime and robust to environment destractions, perturbations etc.

Admittedly this later part is still limited in that the problem needs to be fully observable at all times by the current robot -- it can't deal w say occlusions for it has no memory of what it previously saw -- and its not actually trained to be robust to disturbances and distractions -- what it demonstrated on that front is essentially just what it gets for free from the robustness of deep approaches -- so it could be improved.

WilliamDhalgren··on VR on steam grew 0.06% in july and 0.02% in august
I think compute is more limiting to the graphics than mere pixel density. Certainly if what bothers you is that it seems low-poly.

Now if you can track one's fovea fast enough to render just where you're looking at, then you don't need anything as far off like 8K+; your eye is not nearely that dense.

You do still need a screen with such a density though...

WilliamDhalgren··on Google’s Quantum Dream May Be Just Around the Corner
I guess its quite obvious by now, but sadly I couldn't dig up the video anymore. apologies.
WilliamDhalgren··on Tor’s Branding Pivot Is Going to Get Someone Killed
yup. Though apparently they could deanonymize some folks, but couldn't target the users they wanted.

But I'd expect that to change, a low-latency network like tor that also doesn't create cover traffic just architecturally isn't equipped to deal with something approaching a global passive adversary. Though networks for even that can be built..

WilliamDhalgren··on Tor’s Branding Pivot Is Going to Get Someone Killed
of course; but that's not tor's threat model anyhow. Nevermind the scenario of being mostly taken over; it also obviously falls to traffic analysis by a global passive observer.

In other words, its not gonna protect you from the NSA, ever, nor was it meant to. It might have a fightning chance against the FBI though. Actually it was surprising how in Snowden's files, the NSA was having trouble reliably deanonymizing targeted tor users.

WilliamDhalgren··on Google’s Quantum Dream May Be Just Around the Corner
when they really get there, all current public crypto is just broken. Symmetric is just weakened; basically you need to double the keysize. And there's prob a backlog of messages those actors can then decrypt.

Besides that, everyone needs to move to post-quantum public key cryptographic algorithms. Seems there's usable methods already now, except the signatures will be enormous untill some other, better post-quantum methods are proven safe.

~~Here's an overview of the SOTA there from 32c3 ; I was surprised how usable it already seems: https://www.youtube.com/watch?v=424LHQQB2DE ~~ EDIT: shit, sry that's not the correct video; I'll find it in a sec, I hope..

Or, if you're a bank or a state or otherwise can have a direct 1hop link to your communication partner, you can just use quantum cryptography for extra, physical assurances.

WilliamDhalgren··on Contemplating the possible retirement of Apache OpenOffice
It'd be a very interesting argument, if you could provide statistics to back up the notion that Office products today are really as marginal as you make it out to be. I'd expect them to be comparably central to corporate and civil service jobs in the economy in total as ever, just because economy in bulk is a very sluggish thing and they do the job, but I'd be thrilled to be wrong.
WilliamDhalgren··on Victory for Net Neutrality in Europe
hmm, wiki mentiones that private copies are actually usually allowed in the EU, with the exception of the UK, and a regular reason given for blank media levies many countries have. But that's rather different from online distribution of one's copies.

I think they have some notion of "commercial scale" filesharing to prosecute things like pirate bay, not sure how that works legally. But that too is a different matter, as they don't even share any copies of anything, just metadata, and lately not even that, but merely links.

The real question is whether any of this has any relevance for "pirated" media streaming or downloading being actually legal; not tolerated or thought too minor to warrant the privacy violation of revealing the person behind the IP or any loophole like that, but actually legal.

Honestly, I'd love to hear more about this.

WilliamDhalgren··on Victory for Net Neutrality in Europe
prime-time congestion is hardly exceptional then, so that shouldn't qualify.
WilliamDhalgren··on Victory for Net Neutrality in Europe
copyright law is in fact harmonized on EU level, by the EU copyright directive. Not that it could be substantially different already by Berne convention on a global level.
WilliamDhalgren··on Victory for Net Neutrality in Europe
Well for one, torrent's not a video format. Let alone a streaming video service.

And anyhow, FCC did exactly that; they went after Comcast for throttling torrents for no reason. And throttling torrents just because they're torrents is clearly not permissable by this clause by BEREC either; all it allows for is temporary and exceptional traffic shaping in case of congestion, provided it's done nondiscriminatorily. Which is a good thing; an ISP needs to have a neutral course of action available then besides a meltdown of its net. And yeah, I'd say latency insensitive bulk downloads would be perfectly sensible to deprioritize then. Now, doing what's reserved for recovery in general instead is certainly an abuse of network neutrality, by this rulebook as by FCCs.

WilliamDhalgren··on Victory for Net Neutrality in Europe
Sry, could you clarify, what looks more like rerouting a large download?? Anyhow, allowing traffic shaping in congestion control is not a network neutrality loophole in itself, but ofc any regulation depends on the regulator.
WilliamDhalgren··on Victory for Net Neutrality in Europe
yeah, that's the case. Could you clarify? I'm just seeing old articles claiming the last court verdict on the matter was that the FCC overstepped its authority in that particular case? Did some case or law change this? But yeah, hard to see how discriminating a particular protocol w/o cause could be network neutral.
WilliamDhalgren··on Victory for Net Neutrality in Europe
I wonder even about that, given that it's stated that it can be done only under exceptional and temporary conditions of congestion. The regulator should probably act if it were the general practice instead.

But yeah, the public will need to stay vigilant over how national regulators implement this ruling case by case.

WilliamDhalgren··on Victory for Net Neutrality in Europe
comcast could in the US, and won in court. Despite US regulation being seen as quite robust.

By this text, they'd need to show it was an exceptional and temporary necessity to deal with congestion.

WilliamDhalgren··on Victory for Net Neutrality in Europe
Nope; that would simply be challenged in court.
WilliamDhalgren··on Victory for Net Neutrality in Europe
A legislative act is not an act of the court - that'd be a judicial act - but of the parliament ie a law; that's the legislative body. Ofc a regulation needs to comply with laws; how else is a legal state supposed to work? And afaik traffic shaping done in a content-neutral way is network neutral; for it favors no particular provider. "Provided that equivalent categories of traffic are treated equally" is quite limiting.
WilliamDhalgren··on Why Electric Cars Will Be Here Sooner Than You Think
yeah, plug-in hybrids seem like the best of both worlds; that could still let you go all-electric for the ~90% of your commutes, but w/o the range anxiety or limits to your occasional longer trips. And the fact you end up occasionally driving it like a (hyper-efficient) gasoline car is a totally negligable quantity of emissions in your overall driving.

It's quite possible you can have an all-electric range-extender with over 1000 miles; there was some talk Tesla was looking into current-gen Aluminium-air batteries for that purpose; afaik Al-air is currently the most mature metal-air battery with some actual deployment in limited applications. Downside for conventional aluminium-air is they can't just be recharged once you use up the charge, but the Aluminium needs to be mechanically replaced after the charge has been used up.

Practicalities of that exact scheme sound dubious, but the concept is promising; early gen of some metal-air batteries, even if rechargable could be quite limited in the number of cycles it can last, so be a poor fit for being the main battery bank of an EV car. As a range extender though, this is much less limiting, so could find first use like that, in practice giving 1000 miles range for occasional long trips.

WilliamDhalgren··on Why Electric Cars Will Be Here Sooner Than You Think
really? I thought the superfast charging stations could charge a something like the first half of your battery in something like that time? Granted, gets rather slow as you approach 100%..
WilliamDhalgren··on Planet Found in Habitable Zone Around Nearest Star
actually, the odd thing is how seemingly quickly abiogenesis seems to have occured on earth; few 100 million years give or take. This may mean that given proper planetary conditions, life just tends to happen rather easily - ie for a quite sizablle fraction of candidates at least. Or maybe that it's not going to happen at all unless it happens improbably quickly.

Basically we can't say a much about what the fraction of planets that develop life to planets that have conditions for developing life is based on just one example, but there's no apriori reason for particular pessimism about that number just yet.

Now how likely the conditions are to begin with is a different matter. They've certainly improved in the last decade, since planets at least seem to readily form and earth-sized things too, and apparently easily end up in orbits that could allow for water to exist etc. There's clearly a bunch of other potential conditions that may be necessary for abiogenesis -- untill we know more about how exactly it happens, we can't tell (and actually finding separate origins of life would greatly help in undestanding what that is)

I for one think it very likely this particular planet isn't habitable, because Proxima is such a lousy star, but it's certainly not ruled out. But also that microbes are likely reasonably common in the universe. Around a small fraction of the total number of stars, only some small percentage of what we'd consider candidates today based on their sizes and orbits -- but still that leaves immense numbers.

WilliamDhalgren··on Planet Found in Habitable Zone Around Nearest Star
well, you need rotation to create a dynamo, and these rotate rather slowly; once per orbit only. Now, it does seem this is still quite fast enough to produce a dynamo, but still the issue is whether it can still be strong enough to create a sufficient magnetosphere to save the atmosphere from completely eroding away, given both how close and how active Proxima Cent. is . Here's a paper that tries to see how much atmosphere planets in habitable zones around red dwarfs could lose given the activity of the star and the magnetosphere they could produce: - http://arxiv.org/abs/1204.0275 . I guess its quite possible various other models have different takes on it, but anyhow that's the worry. I hope we'll see such modelling work done for Proxima Cent. b soon.
WilliamDhalgren··on Planet Found in Habitable Zone Around Nearest Star
yeah, kinda true -- though there's a great deal of extra uncertainty about M-dwarfs in particular in comparison to more sun-like stars, just because we did have this one example to study of that.

But on the other hand, we see all kinds of planet types around all kinds of different stars that pose similar problems (say superearth habitability), so even that only applies to questions about planets more similar so some of the types we can find in our own solar system.

WilliamDhalgren··on Planet Found in Habitable Zone Around Nearest Star
Its a really really tiny star; quite small even as M-dwarfs go. So its habitable zone cannot but be where it could tidally lock the planet. Now, maybe some other resonances are possible, perhaps if there are other bigger planets around it helps (and I think they didn't exclude something up to a size of Neptune at about 1AU or beyond); don't know how the orbital mechanics of all this work out.

But there were various modelling attempts to see how a tidal lock affects the climate, and it may not be a dealbreaker; not too thick an atmosphere could effectively transport that heat around, at least so that it doesn't risk freezing the atmosphere on the dark side and making it uninhabitable. But its just modelling so..

Maybe even more problematic is that a tidal lock could imply no internal dynamo, and so no protective magnetic field, and Proxima flares quite a lot, so it could have totally eroded its atmosphere and sterilized the surface with UV etc. Plus Proxima was much much hotter early on; would that period have permanently made the planet uninhabitable?

We just know so very little about habitability of M-dwarfs, and particullary the really small ones; we'll just have to find out,

Before we took a close look with probes, we thought Mars had lichens living on it, to explain its color variations or some such; there's no reason to think we have any more clue about M-dwarf planets than we had about mars then, for we have yet so study any.

Good thing one just happens to be in our backyard :D Though realistically, perhaps it turns out easier to study more distant M-dwarf planets, because this one seems not to transit from our perspective. And I think we could study the spectra and hence atmospheric composition of transiting planets sooner than we can hope for direct imaging of them. Not perfectly sure though.

WilliamDhalgren··on Planet Found in Habitable Zone Around Nearest Star
heh, as you say; apparently there was a previous signal; reading the article about the find at centauri dreams, http://www.centauri-dreams.org/?p=36210 - and it does mention a pre-2016 Doppler signal with a period of 11.2 days around specifically Proxima, that was unconfirmed, and that this mission was to investigate that more closely. "The HARPS Pale Red Dot campaign was created to confirm or refute this 11.2-day signal"

I'm just not sure if that's what the speaker in your video was talking about, of if the embarrasing mistake that was actually announced was the Alpha Centauri Bb planet (or maybe there were other candidates I didn't hear about); I'm gonna watch the lecture you linked and see. Thx for an interesting link, btw!

WilliamDhalgren··on Planet Found in Habitable Zone Around Nearest Star
If you mean the previous claim of there being an Alpha Centauri Bb planet, later retracted, then this new finding involves a completely different star, not just a different period.

This is around Proxima Centauri, the previous one was around Alpha Centauri B. Proxima is a really really small red dwarf, just 12.3% the mass of sun, while Alpha Centauri B is a K-type star, ie somewhat smaller than sun but not that much; around 90% of Sun's mass (And there's also Alpha Centauri A, 10% more massive than the sun). We're not even 100% sure that Proxima is part of the Alpha Centauri system, though I gather it's considered highly likely - it's pretty distant from the Alpha Centauri system, some 15 000 AU (1 AU = distance between Sun and Earth), almost a quarter of a light year.

← PreviousPage 3 of 3