WilliamDhalgren

that's an optimistic way to frame the situation; there's heavy opposition from the content industry to limits to geoblocking, and unsurprisingly the industry seems to have support within the commission (Oettinger at least - perhaps the fact he now left the Digital Economy and Society position helps).

I'm not confident at all they'll be able to completely ban geoblocking in one go. Hopeful they will at least poke some holes in it this round. Most parts of the single market needed a couple of revisions of a directive (quite a few years removed) before liberalizing a particular market fully..

curious: what kind of txt editor(s) would make sense for non-code writing, say a novel, say w markdown or something similar? Especially if the person writing is non-technical so stuff like emacs, vim, atom don't seem like a great fit? I'm thinking something non-obtrusive and with limited options, so that UI is not getting in the way of writing, but with very litte learning curve?

And, given the example of GeorgeRRMartin, perhaps something WordStar-like would fit the bill? If so, are there good free editors in the style of WordStar?

I have this friend, and I've seen the mess of unintended font changes, sizes, styles, bulleted lists, indentation changes. and all kinds of horrible stuff in his manuscripts when I had to repair some of that damage because it was getting unusable - think he'd appreciate a more focused alternative..

sounds like General Game Playing task. Well, MCTS alone is often used in that domain, and it wouldn't surprise me if Ke Jie was a weaker checkers player than just a brute MCTS, nevermind any attempts to use neural nets (if that would even make sense for checkers?)

Hm, well you are no doubt right that it doesn't generalize well to a change of rules. Reminds me of that game DeepZen played. It was trained with a komi of 7.5 and it played too soft and lost when in the actual match komi was 6.5 (or maybe it was the other way around?). A human does not have much trouble adapting to such small rules variations, but at least the version of DeepZen that played that match was hard-coded for that exact komi value, because that's what what used in all of its training examples, and wasn't given as a parameter. It shouldn't be a hard limit of the approach - indeed I think AyaMC was said to have been trained with some flexibility in its komi.

Still, I think AlphaGo does demonstrate amazing positional judgement in unseen board states, and that this is visible in the details of how it plays out particular situations. No two games are exactly alike - difficulty of go for computers is precisely in its extreme combinatorial explosion - and in particular tactical situations every detail of the situation matters. Yet you can see AlphaGo judging the correct sequences of moves, "knowing" how to make a particular group alive for eg, even when a particular other move seems more natural. And probably the most amazing thing about how it plays is how early it becomes completely sure that its got an advantage on the board, and how precisely it judges how much it needs to keep the advantage to the end. Every detail of the board is again relevant here, and basically no human would be so confident so soon. A go bot that couldn't adapt its tactics to unseen situation would be easy to beat; just ensnare it in a large complicated fight, and you're going to kill a big group and guarantee a win. Ofc people tried this in some masterP games, and turns out AlphaGo is tactically just as strong.

So, its basically like with other generalizations you can get from machine learning; a net trained on say ImageNet will generalize to different poses, occlusions, contexts and variations of objects similar to what it was exposed in training etc and still do a superhuman job of classifying such pictures, but will naturally be quite hopeless with completely unseen items. So too AlphaGo seems to know the game of go, generalizing from seen examples to correct judgements in other states, but would be quite hopeless if tested on even a slight variation of the game rules.

> AlphaGo essentially baked good movies into value and policy network by playing millions of times.

I don't think that's a very good description of how AlphaGo was trained at all; you're essentially saying it merely overfits the training set, yet it clearly generalizes rather well to unseen board situations and still evaluates them sucessfully. No machine learning system would be found usefull if all it could do is merely memorize the training data.

Re the use of deep reinforcement learning, well for one the role of reinforcement learning in the first version of AlphaGo, the one described in the Nature paper was rather limited, and a small part of its training; it just made a ~3d KGS policy network into a ~5d KGS bot, and used to generate a training sample for the value net. If we had enough recorded human games to train the value net directly, that'd be an unnecessary step anyhow. And you could create such a training set w/o reinforcement learning since there are pure monte carlo bots stronger than 5d KGS - but that'd be far more computationally expensive.

But its still not really true that there aren't obvious applications of deep reinforcement learning - indeed robotics is one promising application, and that seems rather relevant. this paper initially demonstrated an impressive improvement in manipulative tasks, and you can prob follow its numerous citations for newer stuff: http://arxiv.org/abs/1504.00702

I do agree that this exact architecture in AlphaGo prob doesn't have applications beyond just teaching us how to play go better; it seems too specialized. I believe they mean it in just the vaguest possible sense; that the kind of deep algorithms demonstrating incredible performance in AlphaGo have diverse applications; but this should not come as a surprise to anyone even loosely following what people have done with deep learning in the past couple of years anyhow.

but humans obtain their watts very very inefficiently, so there's prob at least an order of magnitude to give for the same kind of system-level efficiency. Consider a field to feed a human vs a PV installation of the same physical size. And ofc there's all the other ways to obtain electricity...

yup indeed; and also it started already in December, not just January as stated above. Should've double-checked my memory before writing stuff...

It absolutely should be dominant in a long game too. Even if it loses some of its strength at such time settings, it shouldn't lose THAT much, it was just too superhuman. The play should be interesting though; Ke Jie both had access to other strong bots in China for a long time, and could study the records of the MasterP games; maybe he tries something interesting and gets interesting responses so we all learn a bit about the nature of go (haven't watched the recording of this game yet, just woke up).

There was a computer bot championship recently (UEC cup), but AlphaGo declined to participate. FineArts won, DeepZen was second. Think there's a few other chinese bots that could be stronger than Zen but didn't participate. So the real competition didn't bother to show up really.

Nono, you missed the REAL computer supremacy event then; it was the 50ish (!!!) games MasterP bot played in january against the field of top go professionals on some asian go servers. The bot went 50-0, crushing all opponents often in interesting ways.

FineArt is among the bots that have a positive score against top professionals, yes. But it also can lose to them too. MasterP showed that a computer can completely outclass humans!

After the series of games, it was revealed that MasterP is in fact AlphaGo. As far as we can tell from that series, AlphaGo is some serious ELO above other strong bots. So now the question remains - is it that dominant at longer time controls too, as those games were all quick. So that's this match.

> There are quite a small number of old martial art that's pretty decent still.

could you mention a few more? And distinguish if its possible which of those were generally practiced as highly competitive full contact sports (within their rulesets of forbidden strikes etc), and which mostly as a martial art?

hey, that's not bad; didn't even expect SGU had such detailed transcripts.

But actually looking for it now, the particular study I had in mind seems to be from a science or fiction segment from this episode:

http://media.libsyn.com/media/skepticsguide/skepticast2014-1...

Interesting to see how Steve's assesment of modafinil evolved between the two episodes

ah, I'm sorry but I really don't remember. this was some years back.

could be, I really don't know. I can imagine perfectly benign scenarios too; if our long-term memory grows with time, maybe there's just more possible connections/associations to filter out with time too, so that decision just becomes harder, with only the deeper old age being simply tissue decline, the exhaustion of cognitive reserves etc.

Somewhat relatedly, I think I heard on some old Skeptic's guide to the universe episode about an experiment with some nootropic, maybe it was a racetam though I think it was mondafinil, and this kind of reaction time to response accuracy tradeoff was observed, on young healthy adults. Those with less correct answers slowed down, presumably concentrated better and gave better answers, but those already good at whatever the task was simply were slower and yet no better.

yeah, but apparently barely declines untill say 60+. Speed suffers however.

"The adult years were remarkable in that complexity remained at a high level for a protracted period, in spite of a slow decrease of speed during the same period. This suggest that during the adult period, people tend to invest more and more computational time to achieve a stable level of output complexity. Later in life (>70), however, speed stabilizes, while complexity drops in a dramatic way."

and

"These speed-accuracy trade-offs were evident in the adult years, including the turn toward old age. During childhood, however, no similar pattern is discernible. This suggests that aging cannot simply be considered a “regression”, and that CT (completion time) and complexity provide different complementary information. This is again supported by the fact that in the 25–60 year range, where the effect of age is reduced, CT and complexity are uncorrelated (r = −.012, p = .53). These findings add to a rapidly growing literature that views RIG tasks as good measures of complex cognitive abilities [21, for a review]."

well, if only categorisation of extremism online were better. Recently I was searching youtube with the word "Lokiarchaeota", an exciting very recent find in the origin of the eukaryotes, hoping for a scientific lecture, and was mostly getting creationist preachers as results.

Who the hell searches for priests by naming obscure microbes ?? And sadly this is hardly unique; i've been bombarded by UFOs, reptillian aliens and similar outlandish nonsense in search of factual science regularly. And its clearly not doing a reasonable job at predicting my interests at all, for those are not items I'd view.

Why google's imbecilic algorithms promote and push such extremist trash on the general populace is beyond me, degenerating many neutral search results page to worse than reading the yellow press, but this does sadly seem to be what's currently happening, and would reasonably lead many away from wonders that Internet could offer instead.

I don't exactly see what ITS brings to the table here, being a multi-stage rocket as anything else that puts stuff in GTO.

But anyhow yeah - the physics should favor them by practically an order of magnitude, but if they can ever survive to reap any percentage of that potential - they could just as likely just go bust at any further point in r&d, and they don't seem to have either much funding nor are they advancing particularly rapidly...

as far as I understood the starshot discussions, they weren't optimistic about managing to pack any interferometry equipment in such a small package, so there'd be little hope of proper chemical analysis... There was some discussion of what could be figured out from color alone, w/o spectra -- which sounds rather desperate.

A big telescope on earth/in orbit could possibly do more actual work on figuring out stuff about a nearby star than such seconds-long flybys of such wimpy payloads, which is all you get even after you manage the non-trivial challenges of building the phased array, surviving the ISM, establishing communications somehow etc.

at destination? There's no propulsion beam there; all acceleration would have been done at launch, in the vicinity of Earth. You can't point lasers over light-years of distance.

Why? Its supposed to be significantly more efficient, given that its breathing air for launch which enables much higher specific impulse than anything a chemical rocket even could do, and it'd have the operational advantages of being a pure single stage to orbit skyplane, which SpaceX also can't do.

Hard to imagine what else one could even want a launch system to do (except perhaps to be a nuclear turborocket ala http://www.youtube.com/watch?v=C46Dt-X0V8c , if that were politically feasable). And you can hardly say that of anything SpaceX is doing, given that those are simply incremental improvements of existing rocketry tech.

Presuming they ever manage to make it work (which is prob the big question), it should bring significantly more to cheap launch than SpaceX ever could on their current trajectory.

I dont want to go look it up to verify and properly link a citation etc, but I think Schumpeter argued that monopolies are actually good for innovation, at least of the (what's the term?) big bang, fundamental kind, because in a highly competitive marked you can't afford to set aside a significant % of budget for fundamental r&d that's only going to pay off in the very long term and not get outcompeted in the meantime.

The device really should implement speaker identification before accepting commands, yeah. Seems like a task a machine learning focused company like Google really should be able to accomplish.

> The US certainly has many issues, but I find calling it a "Police State", while praising the Islamic Republic of Iran in the same sentence quite ironic.

you mean, praising it as a tourist destination? How's that even controversial? Iran is an absolutely enchanting place to visit.

Now, agreed that calling the US a "police state .. these days" is still quite an exaggeration, but notice the context - the absolutely shocking, police-state-like US visa policy, mentioned in the prior sentence!

You only have to weaken the premise a bit to make it a perfectly sensible claim (and perhaps a bit of rethorical hyperbole is excusable?) : If the authoritarian behavior of the US government forces one to chose between visiting a place as beautiful as Iran, and visiting the US, surely that's an easy choice!

I'm also missing a lot of context here, but I was looking at the leter and she says she's especially sorry she slandered John Sullivan and Ruben Rodriguez.

Wondering exactly how she slandered those two people in particular, I tried googling her name and those two and got some archived mailinglist (and later checked an archived version of her own denounciation of the FSF on the libreboot.org page). She seems to have wanted to see those two people fired. But also another one - Stephen Mahood (who she names "the bully" on the archived page https://web.archive.org/web/20170114100921/https://libreboot... ) . On an archived mail she seems to call both Stephen and Ruben the bullies. https://lists.gnu.org/archive/html/libreboot/2016-09/msg0005...

Given that her heaviest accusations are leveled (also) against Stephen, and that she feels she especially needs to apologise to the other two guys but NOT to Stephen, should we conclude that she still believes at least that guy was guilty of bullying the unnamend transgendered employee of the FSF?

> Does anyone know? Is the claim that this is definitely bacteria, and not Archaea?

Shouldn't really be either; this should be quite a bit older than Last Universal Common Ancestor and hence older than the Bacteria/Archaea split. But I don't actually understand what precisely was found so can't say.

> she was a child by walking to either Spain or Portugal (I forgot which)

how does one walk to Portugal w/o first walking to Spain ? >_>

great post though!

honestly whenever I read anything written about AlphaGo, I wanna start pulling out my hair at the inanity of it. Look at this utterly uninformed comment for example:

> As far as algorithmic ingenuity goes, this is pretty much all there is to it. With all the hype surrounding AlphaGo's victory this year, its success is just as much if not more attributable to data, compute power and infrastructure advancements than algorithmic wizardry.

... Like is anything about this sentence even approximately true??? There was nothing new about the dataset used (and to this day I can't understand why they use the exact datasets they do; KGS for the predictive net, and Tygen for rollout softmax - both amateur player databases, and don't even use the GoGoD database of pro players; seems other teams have comparable or better results at prediction with it), nor in using a sizable cluster to execute a go playing algorithm (limits to the size of it were and are when the algorithm one uses hits steep diminishing returns but anyhow, already MoGo was using rather big ones), nor in training on such a dataset for go playing. The only damn difference were PRECISELY the algorithms used, exactly in contradiction to the claim here! The nugget of truth there is that the algorithms aren't terribly innovative , just their application in this problem - but its training setup and targets are however literally groundbreaking.

Rewind the time a little bit, to the end of 2014, and you'll see the result of an Oxford team, as well from Google's team, that demonstrate a large improvement in the accuracy of the move predictor task - given a board position, predict the next move from an actual game, by using a convnet to do it. That was a first sign that deep learning had potential in go, though at that point it didn't make for a particularly strong player. Systems were getting to a point where they could bias the search effectively with such a convnet for decent gains when the AlphaGo result was announced, that dwarfed these already exciting advances!

So clearly its not about "just" using more computers nor bigger datasets, nor just doing the straightforward thing in applying deep learning to the problem; all of the above was done before AlphaGo, yeah it helped but there was just no contest between the 7d KGS amateur rank the best of the rest were getting and at or beyond top humans AlphaGo did.

AlphaGo came up with the second component, and actually solved a problem thought unfeasable in the computer go world - creating an evaluation function for go. The entire monte carlo tree search revolution of the mid-to-late 00' in computer go was how to sidestep the problem of evaluating if a particular board position is good or bad, by just running full stochastic playthroughs of the game and scoring them instead. AlphaGo on the other hand first created a decent-ish player network (though honestly nothing special - 5d KGS - that's the one that's finetuned by reinforcement learning and notably isn't even a part of the final configuration but just generates a large dataset effectively, cuz humans just haven't played enough games in history for this training setup), then generated a large dataset of games this network was made to play, and then trained (supervised!) a net on predicting the game outcome, given a board position (and taking just one position from each game, somewhat conservatively avoiding overfitting this way).

THIS is the genious of the AlphaGo algorithm; it is a monte carlo tree search algorithm, biased by a convnet, rolled out by a softmax, and crucially with an evaluation function that is mixed, 50%-50% with rollout scores

THAT is a completely novel algorithm, nothing in the literature is particularly like it, in particular the evaluation function and its mix with rollouts was not considered possible! And it works orders of magnitude better than any other monte carlo tree search tried since 2006, as well as orders of magnitude better than other deep learning biased approaches tried since 2014.

And to think that the least of these things; ie the from all the work done since 2014 on AlphaGo, 3 days (!!!) spend on finetuning(!!!) the prediction net by reinforcement learning to make it stronger (though still just a 5d amateur) so as to generate the needed learningset for another component is the only thing mentioned about their approach !?

heh most things are, but I cannot see what complication you may have in mind relevant here.

Vedas don't have any authority in buddhism, just like in other shramana religions, and seem their claim to traditional authority is directly criticized in actual buddhist scripture - literally that this tradition is blind men leading the blind (itself possibly an Upanishadic reference). No doubt there's a decent amount of complication in the matter of various lines of theological influences going between the religions as one would expect, but how is that any justification for calling buddhism a vedic religion? Hindu authors have commonly identified precisely Buddhism, Jainism, Ajivika and the Charvakas, in other words the shramana traditions as being nastika on exactly the basis of their rejection of vedic authority. Now if you'd call it Dharmic religions, sure that works - though is a bit of a strange euphemism in case of Japan where clearly one can only be talking about Buddhism, so why define it in any plurality like "dharmic religions"?

http://www.accesstoinsight.org/tipitaka/mn/mn.095x.than.html

" Suppose there were a row of blind men, each holding on to the one in front of him: the first one doesn't see, the middle one doesn't see, the last one doesn't see. In the same way, the statement of the brahmans turns out to be a row of blind men, as it were: the first one doesn't see, the middle one doesn't see, the last one doesn't see. So what do you think, Bharadvaja: this being the case, doesn't the conviction of the brahmans turn out to be groundless?"

afaik there was never any influence of hinduism on Japan at all - barely a contact with it before late 19.century and modern ways of migration.

And it would be the height of irony to call buddhism a vedic religion, as it along with other sramana movements (such as jainism or ajivika) of the time precisely rejected the vedas.