While we're on the subject, here's a nice two-part article discussing the future AI I was linked to yesterday: http://waitbutwhy.com/2015/01/artificial-intelligence-revolution-1.html

Well that guy on the first image look about 20 so there is 20 years of learning there.

We don't just get born then kick balls we can't do much at first.

Unlike everyone else here, my day job is working in a neurology lab that studies human motion. That means what I'm about to say actually has some relevance to scientific work in the domain. Not to be a dick about this, but watching everyday people talk about "brain stuff" is like watching hackers on CSI: Cyber. It's kinda cute and sounds cool, but has no relevance to reality.

Things that are nonsense: "instinct mode", "muscle memory", and anything to do with neural network models or other AI models in computers. Quantum effects are doubly nonsense, proposed by Roger Penrose for no discernible reason and to the great chagrin of everybody who actually studies this for a living. While there are evolutionary aspects that arguably differentiate us from other animals, they're not particularly relevant here. Animals are perfectly capable of this sort of task at a mental level, but may not express it in the same ways for a variety of reasons.

As I am not myself a neurologist, but merely work with them every day, I will avoid embarrassing myself with any attempt at biological explanations and just give a short high level overview. The heart of it is that scientists do not know much, it's not well studied, and getting together large enough numbers of people to test with is challenging. Even more challenging is producing tasks to test people with that are actually somehow related to human movement in very day life, which is why in our lab we're building a variety of video games to try and tackle that challenge.

That said, there are several key observations. Human motion is enormously complex, from a mathematical standpoint. The amount of physics that need to be computed for multiple joints taking a vast number of variables into account for something as simple as picking up a cup off a table is staggering. We are really good at this, for the most part, having a strong internal model for how the world around us works. Combined with proprioception (the sense that tells you where your body is in space), we are able to solve totally arbitrary problems of motion and control, of sufficient complexity to make robotics experts cry. However, figuring out completely new motions is actually pretty hard. This is a big part of the challenge in sports, dancing, martial arts, etc. You have to think very carefully about how you move in those tasks, for a long time. There are also ways to trick either the proprioceptive sense or the physical model in your mind.

Two things seem fairly clear. One: we have the innate ability to learn, cache, reuse, and adapt motions that we already know. (This is similar to the popular but bogus "muscle memory" concept. Other brain functions behave very similarly.) Two: Our motion system is fundamentally hierarchical. That is, we don't guide our torso, shoulder, elbow, wrist, and fingers individually when picking a cup up off a table - we have a high level command of "pick up that cup" and the details are worked out downstream. While we can take over fine control at a conscious level, it's not the norm.

That cache/reuse/adaptation system is what drives most of our day to day adult lives. As children and throughout maturation, we learn how to manipulate our bodies in different ways and build up a wide library of potential choices. When we're asked to do something completely new, e.g. martial arts, we don't solve these problems up from scratch. We take pieces of existing motions we already know, and stitch them up together from nothing. A child, on the other hand, won't tend to do that because they don't have much of a library or much ability to adapt their motions. Instead they'll start completely from scratch, training completely fresh neural circuits to deal with the problems. This is similar to the difference between your native language(s) and languages you learn later in life - it works exactly the same way for motor control. When speaking French, I'm running an internal translator into English to do everything, not thinking directly in French. As adults, our motions are mapped into existing internal systems and not learned fresh. The problem of what to do when these cached motions are damaged by brain injury - trauma, stroke, disease, etc - is an outstanding and critical one in the field.

Back to your example - a child can't do what you've linked. Lots of videos of kids falling flat on their face or missing hilariously in attempting to do a task like this. They don't have a strong enough internal model of the world or their bodies, and they don't have existing motions to fall back on.

You're right to be amazed by something that everyone takes for granted. Our ability to synthesize and adapt motions that we've learned in the course of growing up is incredible. The scientific work on how we accomplish it is sadly minimal, due to the myriad practical factors that make scientific study hard in general.

Currently we are creating server farms to accurately mimic mouse brain.

It might be interesting watch for you.

On a strictly personal note, I feel that this approach is crackpot nonsense, as far as any relevance to biology or neurology. Probably a fun way to play computer science games and get funding. I get the sense, though, that they think simply building a simulated network based on a vague understanding of the chemical processes involved and letting it go will somehow land up as a viable proxy for actual biological processes. That will teach us precisely nothing about actual organisms, mouse or otherwise. it's just a computer scientist's lab toy.

it's just a computer scientist's lab toy.

yeah, but it's a really cool toy. :)

Thanks for the other information, it was really interesting.

probably none of it is 'instinct'

Humans have barely ANY 'instincts' (see its definition) one might be : a baby seeking to nurse from its mother

Reflexes (ie- when you direct your foot to a place the muscles in the entire body adjust to support the weight and balance the whole through the process)

and learned behaviors (you directing at a high level consciously and unconsciously doing many or the required motions previous learned action... coordination between lower brain upper brain with sensory processing etc.. ).

I don't think the brain does do this automatically. I think it is just because humans kick things around a lot and learn to judge the distance until it becomes second nature (muscle memory ?).

It can only be attributed to muscle memory if you are performing the exact same task. If you change the distance to the ball, but can still kick it perfectly, then it isn't muscle memory. Likewise, many people can accurately throw an object at a target even if they've never attempted that throw before. Again, this isn't muscle memory.

Kicking a ball is performing the exact same task, it doesn't matter what distance the ball is at you will have done the kick thousands of times before and you will already have learned to judge distance. Many people who have never kicked a ball can try and miss whilst people who have never thrown an object at a target will miss.

After reading this good conversation I just can't help but to post another video. Instinct:

Not to be a dick about this, but watching everyday people talk about "brain stuff" is like watching hackers on CSI: Cyber. It's kinda cute and sounds cool, but has no relevance to reality.

Quest failed. A good indicator that you actually will come across as a dick about it is feeling a need to start your sentence with "Not to be a dick". Better to not write the sentence at all then, unless you really do want to be a dick about it.

he heart of it is that scientists do not know much, it's not well studied, and getting together large enough numbers of people to test with is challenging.

Human motion is enormously complex, from a mathematical standpoint. The amount of physics that need to be computed for multiple joints taking a vast number of variables into account for something as simple as picking up a cup off a table is staggering.

And this is the reason us plebs are amazed about the subject, and a good reason to cut non-professionals discussing it a bit of slack.

One: we have the innate ability to learn, cache, reuse, and adapt motions that we already know. (This is similar to the popular but bogus "muscle memory" concept. Other brain functions behave very similarly.)

So Muscle memory then.