7
$\begingroup$

I am concerned by the fact that babies cant walk because the muscles in their limbs arent developed and tuned to give directional control, it takes years before babies gain mobility and dexterity. So technically the neural nets in the motor cortex requires years of training before that capability is achieved.

However, when the babies are born they dont need training for ocular muscles for directional control of their eyeballs (which is quite computationally challenging problem due to high degree of freedom compared to hinge movement) . Do biological neural network for occular muscles come pre-trained while the neural network for limb muscles come untrained by birth?

Improve this question
$\endgroup$
8
  • $\begingroup$ well nice observation, are you saying that rotational or ball-joint mechanisms are hard to tune then hinge mechanisms in the limb? Do you have reference for this? $\endgroup$
    – user0193
    Commented Oct 17, 2020 at 21:31
  • $\begingroup$ @JhonnyS yes, a hinge mechanism gives rise to 2 degrees of freedom (you move in planner axis mainly), while eyeballs move with 3 degrees of freedom (non-planner) so the optimization space is bigger. $\endgroup$
    – gfdsal
    Commented Oct 18, 2020 at 11:55
  • $\begingroup$ That is not my understanding of how eyes develop. Would you care to cite that? $\endgroup$
    – Azor Ahai -him-
    Commented Oct 18, 2020 at 20:43
  • 3
    $\begingroup$ I think you need to get over this computer science idea of "pretrained neural networks" and "computational intensity", because what is going on is far, far more complex, and the brain is not a CPU. There are no "degrees of freedom" issues when comparing the movement of an eyeball with that of an elbow, it's a matter of controlling the muscles that are attached. $\endgroup$
    – MattDMo
    Commented Oct 18, 2020 at 22:29
  • 1
    $\begingroup$ @MattDMo I think he is right, and there is no contradiction to your view. The computer science side is far, far more complex, if it is needed, and with a neural network, the CPU is possibly not used except for simple unrelated tasks like handling data. The aspect of degrees of freedom is expressed in the muscles needed. $\endgroup$
    – Volker Siegel
    Commented Oct 19, 2020 at 6:23

3 Answers 3

Reset to default
8
$\begingroup$

Biological neurons function in a very different way, as compared to the simplistic artificial neural networks of machine learning. For example, see how real neurons work and how they connect with each other. The types of neurons themselves are very varied: "...neurons to take specialized forms such as unipolar,bipolar, multipolar, anaxonic, pseudounipolar, basket cells, purkinje cells, Lugaro cells, spindle cells and more.".
As for the eyes, light passes through a large network of neurons before hitting the rod and cone cells. It's the opposite for the octopus retina. Animal babies are able to learn to walk much quicker than human babies. Also, have a look at videos of how blind children learn to walk.

So essentially, there's a lot lot lot more going on in the creation and working of the human body than just "neural network training". A really nice paper titled "facts and anomalies" gives perspective into this complexity.

Nobody fully understands how the system of cells and the equilibrium works. It's extremely complex and multi-dimensional. When you look at the complexity inside a single cell and the electron transport chain, you'll notice that there's a heck of a lot of things that are already pre-programmed and designed to make use of the molecular properties of nature to create structures and motors using the equilibrium of chemical properties. When such an equilibrium is created, evolution and adaptation can happen, but only in a very controlled manner. So I do not believe a baby's neural network is getting trained to walk or see. It's pre-trained from millions of years of evolution. The reason it may be taking some time to mature, is probably because the corresponding neurons in the brain are taking some time to get created, grow and find their way in the brain to form the right synapses.

Improve this answer
$\endgroup$
6
  • $\begingroup$ So I guess what you are concluding that in human babies, its the combination of molecular properties and pre-trained neural nets and these pre-trained neural nets that happens due to pre-existing information in DNA which got encoded over time, millions of years of evolution? $\endgroup$
    – gfdsal
    Commented Oct 18, 2020 at 12:05
  • 1
    $\begingroup$ @gfdsal: Personally, I do not believe only DNA contains the information. There's a lot more in the cytosol we haven't even discovered. Just have a glance through the "facts and anomalies" paper that I linked to, and you'll see why it's too early for us to make conclusions. $\endgroup$
    – Julian
    Commented Oct 19, 2020 at 14:05
  • $\begingroup$ your answer coupled with @MattDMo 's answer made lots of sense. I am accepting this due to references. $\endgroup$
    – gfdsal
    Commented Oct 19, 2020 at 20:11
  • $\begingroup$ I am just curious why did you mention that light passes through a large network of neurons before hitting the rod and cone cells. isnt this other way round, because rods and cones are the main receptors of light which then passes to visual cortex instead of visual cortex passing the information to rods and cones? $\endgroup$
    – gfdsal
    Commented Oct 19, 2020 at 20:45
  • 1
    $\begingroup$ @gfdsal: Do a bit of Google searching to see and understand the function of the layers of the retina. I didn't mention anything about the visual cortex, but I have reason to believe that data is transferred from the brain to the retina too. Could you please do plenty of Google searches about the questions you have: neuroecology.wordpress.com/2017/03/29/… $\endgroup$
    – Julian
    Commented Oct 22, 2020 at 7:04
7
$\begingroup$

Some babies start walking by 12-15 months, so it's not exactly years before they can do it. They can also grasp things with their hands from birth. When first born they still don't have full ocular control, and especially can't focus except on strongly contrasting objects. If you watch a newborn, their eyes are often moving from place to place, even when a parent's face is in view.

As for why they don't have much muscular control, just think about their circumstances – they've spent the previous 9 months in a very cramped uterus, with barely any room to move around and no chance to fully extend their legs when they're in the third trimester and increasing in size quite rapidly. It's not that their muscular control neural networks aren't trained at all, they're just not trained to their full extent yet. The eyes are a bit more trained, because light can penetrate inside the womb, especially at the latest stages of pregnancy when the mother's abdominal skin is stretched tight, and they have plenty of chances to move their eyes around, even if they have nothing to focus on.

Improve this answer
$\endgroup$
4
  • $\begingroup$ But doesnt child need data to train the cornea muscles? Also are you saying the light can penetrate moms belly onto the womb, but do you think its significant? are there any references for this as I am also curious to know. $\endgroup$
    – user0193
    Commented Oct 17, 2020 at 21:28
  • 2
    $\begingroup$ @JhonnyS yes, light can penetrate skin rather easily. As a simple experiment, stand in front of a mirror in a very dark room, then open your mouth and shine a flashlight into it. You'll see your cheeks light up quite nicely. As far as specific references, I'm not sure - it's in just about every baby book I've read. Sound also penetrates easily, so assuming both parents are living together, the baby already responds to the sound of their father's voice when they're born. $\endgroup$
    – MattDMo
    Commented Oct 17, 2020 at 21:49
  • $\begingroup$ So is the answer that biological neural network in babies come with partially pre-trained networks? So this partial pre-training happens in the womb as you suggested or its due to genetics (half from mother and half from father)? $\endgroup$
    – gfdsal
    Commented Oct 18, 2020 at 12:00
  • 2
    $\begingroup$ @gfdsal Both. There is plenty of information and organization in our brains that is influenced by DNA, and plenty that comes from experience. It's not either/or. Biology is extremely complex, as the answer by Julia points out. We don't understand all the interactions that take place, or how exactly the brain develops. $\endgroup$
    – MattDMo
    Commented Oct 18, 2020 at 22:21
1
$\begingroup$

Walking is about more than having strong legs moving in a hinge motion. It's about coordinating all your leg muscles and your torso muscles. Babies have instincts which tell them how to coordinate all the muscle movements for suckling, but they still have to practice all the muscle movements for eating (and again, while learning to coordinate all the torso muscles for sitting). Eye movements are far simpler.

Improve this answer
$\endgroup$

Not the answer you're looking for? Browse other questions tagged .