The Human Brain Project, Cognitive Computing

[BryanM]

Haven't had any future-fetish transhumanist threads in a while. So.

A brief FAQ:


What the hell is the Human Brain Project?

A while back there was a project dedicated to simulating a neocortical column of a rat on a super computer, the Blue Brain Project. The idea basically being, this was the most complex basic building block of a mammal's brain, so if it was possible to simulate this section, it would prove that creating a full rat brain would be possible.

It was successful enough that it won some 1+ billion euros worth of science funding from the Eurozone for a decade+ project, with the hope of having a working scaffolding of the human brain by 2020-2023ish. Some neurologists were very butthurt over this and there's been some news lately of them whining about it.

According to the project lead, this is worthwhile due to the following assertions:

* There's like 100,000 papers published a year on the brain. It's impossible for any human being to ever be able to benefit from it all. Not without a model that consolidates all of our knowledge.

* We have bum fuck-all on how intelligence, consciousness, all that sort of thing, really works. When will we magically pass the threshold were we'll suddenly know enough? 1,000,000,000 papers? 1,000,000,000,000 papers? Better to build an exposed black box we can look inside.

* It will have benefits in the field of medicine, will contribute to artificial intelligence, and would give better specifications for the kind of computer hardware we'd need for this problem domain.

He compares it to the moon race often, saying we need to have a "brain race".


Is there more information out there on this thing?

In written form, there's pretty scant info out there for us laymen. There's lots of youtube videos with awesome long boring vague lectures. Lots of references to dicing up mice and rat brains. Lots of whining about how getting human data is difficult - cadavers and small chunks of brains removed due to tumors.

I remember there being some kind of data-sharing site, but can't find it... meh. The official web site is more of a sales pitch kind of thing.


Having to wait around 10 years for a little virtual man spasming wildly on the floor, puking and peeing on himself is boring! Can't we speculate wildly in the meantime instead?

YES THAT'S WHAT THIS THREAD IS FOR!!

(Well, unless anyone has some legitimate expertise or info to dump onto it.)

In my own opinion the entire thing comes across as rather Frankenstein. Even without the carving up rodent brains bit. This is a black box way of generating a mind; if successful the brain simulated would be little different from one made of meat. To get anything useful out of it, besides a functioning brainstem, the thing will have to be taught a language. Like in the game A Mind Forever Voyaging, it would need to experience a "childhood" of sorts. A feat itself possibly as difficult as the brain simulation; I'd add another 10 to 15 years to it for The Human Mind Project.

And in the end you end up creating a person. Creepy!

A Truman Show/Big Brother type reality TV show could be used to drum up support and funding. But what really comes to mind is the morality of ever using such biologically-derived blackbox AI in robots - for them to do what YOU want them to do, instead of what they want to do, they would have to be "brain-damaged" in some manner. An inhibitor mechanism that prevents them from killing themselves or other people. And god knows, if this is the only way to get those robot wives everyone keeps talking about



I'd pass.

[Lord Satori]

It's fascinating when reality seems frighteningly similar to science fiction. It seems like this'll be one of those things that will either turn out to be very good or very bad, with little room in the middle.

[DEL]

Well if the elites want to live on in some kinda transhumanist transcendence form, then I'll happily wish that horrible fate on them.

I'd pass too.

[Doctor Butler]

This seems like it might become a polarizing talking-point on pop-media "news" stations (playing-GOD, abominations, etc.), before funding dries up due to a plateau in tangible progress, and it quietly disappears into obscurity.

[BulletMagnet]

Maybe my cynicism is clouding my vision, but it always seems to me that whenever some manner of research or other is done into the inner workings of the human brain, the only people who ever end up getting any notable use out of it are advertisers.

[BryanM]

Some of the other things going on in this topic:


"Turing Test Is Passed" Hoax

Some headlines claimed the Turing Test was passed miraculously and magically. Out of nowhere. Five seconds later it was shown to be a fraud. Journalists impressed everyone with their skill of vapid ignorant bullshitting and wasting everyone's time.

Those people who claim passing the Turing test "doesn't mean anything"... they're wrong and dumb and smell bad. Passing the Turing test is an enormous accomplishment. Look at the type of things it has to endure in just a minute of text exchanges:

Basic English comprehension and object comparison: Which is bigger, an elephant or Neptune?

Basic agency and learning: Let's play guess the number. Nothing says "human" like being able to learn and play a game.

Memory: What animal was it I mentioned earlier?

Eugene could not last ten seconds of "are you a chatbot?" type questions because he is nothing more than a chatbot.

A program that could pass the Turing test, at the very least, would be a little buddy that could play Dungeons and Dragons (and other games) with you. In case you need to level up a Dwarf at 2 am every night for the rest of your life.


Google

These guys hired Ray Kurzweil to lead a department. Ray Kurzweil, the grand wizard of the singularity religion.

They bought a company called DeepMind recently. I have no idea what's going on there.

They also seem to be at the forefront of the automated car, having bought all the people who finished at the top of the Darpa challenges. But of course, ghost cars aren't general cognitive computing, just a solution for a specific domain. A domain that'll render over 10 million people in the United Stated unemployed eventually.


Obama's BRAIN Initiative

This one was announced a few days after the Human Brain Project was announced. It was kind of cute, like the president saw the Europeans doing a thing and he didn't want to be shown up so he went on tv to say "I have a brain thing, too!" But of course these things take years of planning to plot out.

This one might be even more ambitious than the HBP, since they plan to map the brain. At current rates and methods this would cost a trillion billion million dollars and involve dissecting lots of human heads. So obviously some new techniques to get measurements will need to be developed.

[Battlesmurf]

definitely interesting. On one hand- this stuff is neat, and I'm curious if it could be pulled off. On the other hand, if it can be pulled off- how long until Skynet happens?

[Khan]

Ghost in the shell eat your heart out!

[BryanM]

IBM announced a 4,000 core chip a few weeks back. The webpage for it suggests in the future, tricorders from Star Trek could sniff and identify bacteria inside our bodies.

The biggest issue on the hardware side is the issue of power - for something like the Human Brain Project to simulate a human brain with today's hardware it allegedly would require all the output of a nuclear power plant.

These so-called "neuromorphic" cpu's aren't classical cpu's - you can't play Starcraft on them. They're designed to be optimized for different problem domains.

It goes without saying that IBM would like one day to see this kind of chip in every cell phone in the world. Rather a lot of money if they get in there first.

definitely interesting. On one hand- this stuff is neat, and I'm curious if it could be pulled off. On the other hand, if it can be pulled off- how long until Skynet happens?

Intellect seems pretty tied to the number of neurons in the brain and the connections between them; as the man humbles the pig that humbles the mouse that humbles the lobster. It may prove to be nearly as simple as "scaling up". Going from nothing to something would certainly be harder than going from something to something better than human.

Stephan Hawking has been pretty vocal about the danger of this for the past several years now. Here he is a few months ago telling it to a British man on hobo.

The sad comfort is that unless said god has a sense of humor, there will not be metal robot soldiers striding over a field of human skulls. Mop up duty after the nuclear holocaust would be more efficiently carried out by turbo dogs: smaller targets, faster, more efficient to make. Or even more boringly: robot insects that slowly poison people to death.

It's kind of like that time when I learned it'd require less energy to push the Earth into Jupiter in order to destroy it, rather than the Sun. Reality isn't often very exciting, is it?

[Lord Satori]

I don't want to know about the lifeforms that lurk inside me. I don't know about them, and they don't know about me, and I like it that way. I'm pretty sure everyone else feels this way as well, so save the human race and abandon the project now.

[Acid King]

I can't wait for this to become a reality. I'll probably long dead by then, but maybe my progeny can scrape some grey matter from my decaying skull and put my consciousness in an unstoppable killing machine.

The biggest issue on the hardware side is the issue of power - for something like the Human Brain Project to simulate a human brain with today's hardware it allegedly would require all the output of a nuclear power plant.

...and will be so expensive only the five richest kings of Europe will own them.

[Gespenst]

Now why the heck this reminds me of Front Mission. BD device, S-type device, plus someone's orchestrating a war to test out these technologies


Wait. COFFIN system plz

[BryanM]

...and will be so expensive only the five richest kings of Europe will own them.

Professor Frink knows his stuff!

[null1024]

This sort of thing sounds cool. When I was younger, I always used to [okay, I still do] think about things like straight up brain-emulation and the like for AI. And after seeing how processing technology is moving more and more towards extremely parallel systems, it seemed like a natural sort of fit.

that being said, I already feel sorry whatever poor bastard we spawn the first time we get an actual working human brain up and running [to be fair, by the time we do emulate a human brain, and not a different animal, we may have perfected the process], and there's going to be a fuckton of ethical issues about it all, especially if we find ourselves increasing the neuron count to something that isn't quite human, but able to learn and communicate [through language] and all that

Maybe my cynicism is clouding my vision, but it always seems to me that whenever some manner of research or other is done into the inner workings of the human brain, the only people who ever end up getting any notable use out of it are advertisers.

Just imagine!
In the future, we will have vat grown advertisers running around. Instead of a normal brain, they'll use manufactured artificial ones.
Much easier -- instead of training them all, you just image the brains with the mind of one you trained the long way.

[I was going to write something really silly about how they'd also replace us and take over the world -- all so the advertisement firms [already run by manufactured minds] would use the process to make their own consumers that are completely compelled to buy from them or something, I dunno. The idea kind of died as I started to type it out.]

[BulletMagnet]

I don't doubt for a second that someone out there is working on that. And has already seen more success in said endeavor than he should have in a rational world.

[BryanM]

Today the theme will be

Moravec's Paradox

Put shortly and simply, this is the idea that our skills that are hundreds of millions of years old, such as vision and walking around, are vastly more complex than the skills we developed recently, such as throwing projectile weapons or playing chess. Our brains are just really damn good at those old problems than they are at the newer ones.

For example, take the hypothetical Terminator kill bot. Shooting someone is basic geometry with a slight vertical offset based on distance (I calls this arithmetic) to take into account the gravity the projectile will experience. Even the dumbest killbot should be able to hit a standing target in the leg from a sizable distance with not even microseconds worth of computation; something a human would take tens of thousands of hours of practice to even begin to compete with.

While on the other hand, it's really difficult to believe in the future of the robot überman if you skim the latest DARPA Robot Challenge. Even knowing they get no bonus points for speed (these are supposed to be rescue bots, after all), it's disheartening to see them crawl so slow.


IBM Japan Made Some Propaganda

IBM Japan teamed up with the Steins;Gate people to create a few Japanese cartoon shorts about the glorious magical future of poop-sniffing cell phones. That also give fashion advice and are considerate about their owners' massive egos.