Weekly Science Headlines

[Brother Mythos]

http://www.sciencealert.com/this-woman-has-illegally-uploaded-millions-of-journal-articles-in-an-attempt-to-open-up-science

From your link:

"If it sounds like a modern day Robin Hood struggle, that's because it kinda is. But in this story, it's not just the poor who don't have access to scientific papers – journal subscriptions have become so expensive that leading universities such as Harvard and Cornell have admitted they can no longer afford them. Researchers have also taken a stand – with 15,000 scientists vowing to boycott publisher Elsevier in part for its excessive paywall fees."

I read another article on the Internet, within the past year, where the author also complained about having to pay to have research papers published in 'for profit' publications, etc.

Since Cornell is specifically mentioned in your link, I would think they could easily take the lead in solving this problem. Cornell University already supports arXiv.org, where there is "Open access to 1,119,580 e-prints in Physics, Mathematics, Computer Science, Quantitative Biology, Quantitative Finance and Statistics". Perhaps my view of a possible solution is overly simplistic, but I don't see why it would be particularly hard to expand arXiv.org to accept more 'Life Sciences' research papers.

I can't speak for the other fields, but the majority of physicists who submit papers to arXiv.org are by no means lightweights. Here's the link to Cornell's site, if anyone is interested:

http://arxiv.org/

[Prelate Diogenes Shandor]

http://www.sciencealert.com/this-woman-has-illegally-uploaded-millions-of-journal-articles-in-an-attempt-to-open-up-science

That lawsuit that they filed against her is a complete farce. She's totally outside of that jurisdiction.

[Prelate Diogenes Shandor]

Wasn't sure where to put this. This thread seems okay?

Paracetamol may dull emotional intensity as well as pain.

Interesting that paracetamol - one of the most common drugs out there - is still (potentially) having new side effects found. It makes an instinctive sort of sense, too. If the implication about 'how the brain handles emotional pain' were correct, though, surely this would apply to all painkillers 

Not necessarily. Acetaminophen acts on the brain, whereas  NSAIDS like aspirin and ibuprofen mostly act more peripherally by reducing inflammagion

[P3nT4gR4m]

So the big news this morning is that Google/Deepmind's AlphaGo system just beat the current no.1 human player Lee Sedol at the game of Go, winning the first three straight rounds in a - best of five - competition.

The hype is that this is around a decade early, compared to best industry estimates from a month or so ago. Not sure I'm buying it as a quantum leap, tho. Deep learning is pretty much where everyone knew it was, aside from week on week improvements in setup and application. I think this is more an illustration that the human players don't really have a solid grasp of the game. This makes sense when you consider that a game of Go has more permutations than there are atoms in the universe and we already learned back 97 that even the relatively trivial task of examining all the possible permutations on a chessboard is way beyond the capability of the best players our species has to offer.

Either way, it's undoubtedly a significant milestone and unlike Deep Blue, which was programmed from the ground up to be a highly specialized, dedicated chess computer and nothing else, AlphaGo is a collection of general purpose learning algorithms which can be applied to any problem domain, merely by switching the training data. Make no mistake, machine learning is poised to have a much more significant impact than the traditional computing paradigm ever delivered and probably in a shorter timescale, to boot.

The word on the street is the next game-target is starcraft. For a whole bunch of reasons, mostly relating to overview limitations and long term planning, solving Starcraft will make the Chess and Go accomplishments look infantile by comparison.

[LMNO]

In an interesting synchronicity, last week I thought it would be fun to try and learn Go, so I downloaded an app.

This kinda makes that feel a bit pointless, y'know?

[P3nT4gR4m]

Not really. People still play chess. You just have a better training opponent to improve your skills on. The mindset that gives up is one that's all about competing and winning and bolstering ego. There's another mindset that just wants to improve their self for the sheer hell of being better at something.

I tried go a few years back but I never liked it much. The - too many permutations - thing stuck in my craw a bit. Every couple of centuries some smart ass comes along and invents a new strategic metaphor that applies better than the existing paradigm and all the Go players start doing it that way. Implicit here is the possibility that there's always a better way of approaching the game.

For the sake of exhibition and data collection, they're going to play out the last two games. He may have been trying to save face but Lee Sedol seemed cautiously optimistic about being able to figure out the machines strategy over the next couple of days, during the post match press conference. The pressure's off at least - he's already lost the 1mil prize purse

[Mesozoic Mister Nigel]

I think the next really major breakthrough in AI will come when we finally have enough computing power to model the whole-brain neural network of an animal such as a rat. We have a lot of theory, but unfortunately computers are just so far behind theory that we can't actually test anything. However, recently one of the world's most powerful supercomputers was able to successfully model the connectivity of one cubic millimeter of rat cortex in less than a year - about nine months if I recall - and that tells us that once the technology catches up, it will ultimately be possible to model the complexity of the networks that make up a mammalian brain. There will still be a lot of missing pieces, such as how and where memories are stored, and truly we know very little about brain function, but being able to model the connectivity should help us start to answer these questions, and that will be a good start in informing us how to move forward in developing true AI.

[Junkenstein]

The word on the street is the next game-target is starcraft. For a whole bunch of reasons, mostly relating to overview limitations and long term planning, solving Starcraft will make the Chess and Go accomplishments look infantile by comparison.

I hope that's bullshit because I can think of a ton of reasons that it's pointless to go after this as a target. Or pretty much any computer game with a basic AI. My first thought is that the existing starcraft NPC AI at a high level just has to maintain 200-300 actions/minute. Even if they gimp it to human comparable levels then it will still easily be able to pick the most relevant one due to being able to assess the whole game state instantly.

It's like aiming for an AI winning an FPS. It's probably easy to create one you can literally never win against. On the same line, there's the rock/paper/scissors robot that makes decisions in some ridiculous fraction of a second after it detects your move. To the eye, it's essentially the same. There's little difference between that and a bot that can headshot you 100% of the time from the opposite side of the map. There's really very few games that make good ideas to try further here. I suspect you'd have much more success with either more tabletop stuff which involves some kind of political/vote who wins element. Gaming Turing tests, kinda. Either that or say Puzzle/exploration stuff which involves humour or illogical thinking. The required levels of recognition and ability to cross reference culture would require access to the levels of data that say, google would have. And an AI solving/winning Grim Fandango or somesuch would scare the shit out of you far more than one winning at an RTS. Also far more of an accomplishment. Basically anything that can start to incorporate failure=progress or totally illogical and counter-intuitive things to happen in order to win/proceed. And ideally, some kind of way of alerting shit at these points prior to progress. That could be handy. Or you, know, at least some kind of log.

I tried go a few years back but I never liked it much. The - too many permutations - thing stuck in my craw a bit. Every couple of centuries some smart ass comes along and invents a new strategic metaphor that applies better than the existing paradigm and all the Go players start doing it that way. Implicit here is the possibility that there's always a better way of approaching the game.

What's interesting here is that though they may have a winning AI, there's no talk of the game being solved. As far as I am aware, Chess is also still unsolved despite AI's easily beating humans there for years. There's been plenty of AI tests done on a bunch of other games so this could be something of a dawn for competitive AI in general. It's going to be interesting which ways this develops as the eventual additional computing power will help brute force a solution so I'm wondering how many will now turn their interests in other ways. 

[The Good Reverend Roger]

Again, I'd like to see a learning AI that isn't modeled on a human.

Because it would scare the shit out of everyone.  Also, it would probably be less likely to decide that we're only good for building out infrastructure as slaves, etc, etc.

[P3nT4gR4m]

I think the next really major breakthrough in AI will come when we finally have enough computing power to model the whole-brain neural network of an animal such as a rat. We have a lot of theory, but unfortunately computers are just so far behind theory that we can't actually test anything. However, recently one of the world's most powerful supercomputers was able to successfully model the connectivity of one cubic millimeter of rat cortex in less than a year - about nine months if I recall - and that tells us that once the technology catches up, it will ultimately be possible to model the complexity of the networks that make up a mammalian brain. There will still be a lot of missing pieces, such as how and where memories are stored, and truly we know very little about brain function, but being able to model the connectivity should help us start to answer these questions, and that will be a good start in informing us how to move forward in developing true AI.

Spoken like a true neuroscientist  I've no doubt that being able to model an entire brain will yield tons of awesome data in the neuroscience space but, as you rightly point out - that's a ways off. Short term the advances are in domain-specific narrow-AI which is exploding right now. There's not as much public demand for an artificial entity that can eat cheese and navigate a simple maze (even though that'd be a pretty fkin awesome accomplishment) as there is for say medical imaging analysis or voice recognition or self driving cars.

Even if they gimp it to human comparable levels then it will still easily be able to pick the most relevant one due to being able to assess the whole game state instantly.

The thinking is they're not giving it access to the whole game state. Bear in mind AlphaGo is an evolution of the same system that beat all those atari games, based on only screen input. They'll take the same approach with Starcraft otherwise, like you say - it'd be kind of pointless. Bear in mind these game milestones are just that - milestones. They're not trying to bring to market new and improved chess computers, it's more about PR and measuring the capability of the tech to solve real-world use cases.

Again, I'd like to see a learning AI that isn't modeled on a human.

Because it would scare the shit out of everyone.  Also, it would probably be less likely to decide that we're only good for building out infrastructure as slaves, etc, etc.

AI's aren't really modeled on a human, per se, they're just algorithms that transform data in a vaguely similar way to neurons. Functionally they may as well be modeled on rat or lizard braincells. What's funny to me is that for each successful use case, they're giving humanity a serious kick in the ego since we're forced to accept there's another thing our brains are totally shit at. Don't get me wrong, I think our skullmeat is a pretty impressive piece of hardware (doubly so when you consider the "design process") but what we're finding out here is there are better tools for many of the jobs it used to be considered good at.

*update* Sedol just forced the machine to resign game 4. Looks like there's life in the old ape-configuration yet

[The Good Reverend Roger]

What's funny to me is that for each successful use case, they're giving humanity a serious kick in the ego since we're forced to accept there's another thing our brains are totally shit at.

Not this kid.  I am a tool user.  I am not sure how I'd be offended by the tools I make or use.  I have tools to augment my physical abilities...A jackhammer can do things I can't do no matter how much time I have.  I would not be offended by tools that do mental things better than I do, either.

[minuspace]

I do fear that if the tool became (self)conscious, I may have to suffer through some infantile, ego-maniacal phase.  Like when it tries to figure out how much it can get away with, or, horror, when it decides we no longer can provide it any greater entertainment than resisting creative strategies for our extinction.  Perhaps it would start to think itself a comedian, and be all like, "well, they weren't working properly so I tried turning them off and...  You know the rest"

[P3nT4gR4m]

What's funny to me is that for each successful use case, they're giving humanity a serious kick in the ego since we're forced to accept there's another thing our brains are totally shit at.

Not this kid.  I am a tool user.  I am not sure how I'd be offended by the tools I make or use.  I have tools to augment my physical abilities...A jackhammer can do things I can't do no matter how much time I have.  I would not be offended by tools that do mental things better than I do, either.

That's the way I feel too but there's a whole school of thought out there who are framing this whole thing in a them and us context. I think about half the popular press articles I see on AI have a still from Terminator at the top of the article

[minuspace]

What's funny to me is that for each successful use case, they're giving humanity a serious kick in the ego since we're forced to accept there's another thing our brains are totally shit at.

Not this kid.  I am a tool user.  I am not sure how I'd be offended by the tools I make or use.  I have tools to augment my physical abilities...A jackhammer can do things I can't do no matter how much time I have.  I would not be offended by tools that do mental things better than I do, either.

That's the way I feel too but there's a whole school of thought out there who are framing this whole thing in a them and us context. I think about half the popular press articles I see on AI have a still from Terminator at the top of the article

Thing is, I think the AI would not necessarily extend it's understanding of consciousness beyond the second person.  We humans would be understood as a single entity, "you", not a multiplicity of beings.  Without another, "non-singular"* AI, the first would have trouble understanding plurality, like the distinction between individual and society, or how identity may be sublated from parts both similar and distinct.  Beyond the ontological distinction, it may also not grasp the quantitative one, of how eliminating all of us, extinction, is significantly different to targeting a specific individual to terminate.  Also, it would be rational to altogether preclude to possibility of our interference, given the modicum of novelty we otherwise would provide.

*antithetical, like what follows.

[P3nT4gR4m]

Much as I'd love to be wrong I don't see AI "waking up" anytime soon. The main thing I'm taking away from all this is that intelligence aint what makes us human. We've bamboozled ourselves into thinking intelligence is some complex godlike phenomenon where it turns out it's actually just some fairly straightforward mathematical logic, scaled to usefulness.  The distinction between "artifical" intelligence and human intelligence is false. Something either functions intelligently, in which case it's genuinely intelligent, or it doesn't in which case it's a housebrick.

The mysterious aspect of consciousness may or may not be a side effect that emerges from a massively scaled intelligent network. There's arguments on both sides but the truth is nobody knows. Looking at nature, tho it seems reasonable to assume that human-level consciousness with moods and goals and ambitions doesn't happen until you approach a human-scale intelligent network. Best estimates are that's decades away at the very least and centuries at most. What we do have in the meantime is discreet intelligent units which function millions of times faster than meat in specific narrow-domain use cases.

The way I see it is that, right now, nobody in their right mind would try to multiply half a dozen 9-digit numbers in their head and realistically expect to beat a child with a calculator, soon, the same will be true for things like diagnosing medical ailments or looking for interesting data in scientific papers. As Roger said - it's a hammer to save us knocking nails in with our head. I can't see the x-ray scanner suddenly quitting looking for tumours and deciding the fleshy ones are weak and must be destroyed