This is part II of the post that I began a couple of days ago. In the previous post, I covered how human beings remember, and also to a certain extent, how computers work.

So now, lets cover some more complex memories. How do we remember things like say Newtons First Law of Motion? Oh wait, I know this one. “Any body remains in a state of rest or in a state of uniform motion in a straight line unless acted upon by an external force”. It took me a couple of seconds to remember this, but I think it’s a pretty good definition.

The definition of Newtons First Law of Motion is a pretty standardized thing. It involves a sequence of words which are in a very clear sequence and make absolutely no sense if you do not put them in the right order. I can’t remember this if I put it in my splendid data warehouse of my previous post. If i try to remove any pieces of data in this definition in the interests of quicker recall, I will screw up the meaning itself, which just would not do. So how is it done.

Again, the mechanics of it are probably beyond my abilities. Something to do with REM Sleep for all I can tell. But conceptually, we can think of this procedural memory or sequential memory more in terms of our standardised data warehouses that exist today with computers. If a computer were asked to quote Newton’s First Law, I am certain it would do a far better job than your humble blog author here.

This is a part of information retrieval that the computer is inherently better than human beings. We suck at definitions and tests. It took me almost 15 days of slogging 4 hours a day to memorize the key parts of my commerce textbook. And yet, the only part of it that I can remember 14 years later is “Money is a medium, measure, standard and a store.” That is pretty bad, considering I can still do a 10th standard Physics Paper today and expect to at least get 75-80%.

So what makes commerce so hard to remember and physics so easy to? Is it because I use physics every day, and commerce never? Well, that does not wash, because I work in a bank! The closest I get to physics is the advertisement for Physics Tuition’s that I see while standing and waiting for my train to take me to work! In my case, I suspect it’s because I genuinely learnt the concepts behind each of those physics definitions I memorised. So, I understand that uniform motion in a straight line can be expressed as similar to a state of rest in an ideal world, however odd it may seem at first glance.

In someone else’s case, this might be completely different. They might consider the 10 features of a joint stock company to be so logical that they might not be able to understand why I could not remember them. (My mom is one such example). But once I get this insight, the rest of my theory is now obvious. Even in cases where we memorize extremely long sequences of events, long-term recall is ensured by understanding of the basis of such a sequence. Else it would seem exceedingly unlikely that any person would be able to memorize and retain the memories in a loss-less manner unless they either understand the concept/context of the memory perfectly, or if they relive the memory so often that its brought into the intrinsic data warehouse explained in my previous post.

One final example to belabor the point. If someone were to ask me what is 12 times 12 I would be able to answer 144 without a second thought. But thinking about it, the concept of arriving at 12 times 12 is quite laboured. If I tried doing it from first principles, it would take me at least 5 minutes to re-understand the concept, and then to apply it to this specific case of 12×12. This is the fundamental difference between the “intrinsic” memory of the brain, where we store our most our automatic responses, and the “learned” memory of the brain, where we store corollaries, and concepts that don’t need reusing all the time.

Now, this does not really add anything. It’s just a neat (in my opinion) understanding of the way memory works. But the key is if someone were to figure out a way to segregate these memories, build short-term data warehouses that computers could access instantaneously and “guess” answers while they search their complete data banks to get the “right” answer, we would have a pretty neat form of AI. Now, I have not seen Watson, who is supposed to be the computer who is the Jeopardy Champion, but maybe Watson’s AI has this sort of guess algorithm, where if Watson has an answer which is significant to 2 sigma, he presses the buzzer.

I have no clue as to how AI algorithms and machine learning theories work. But if the any of them use the thought process that I have set out, it would be intriguing to see where it would finally end up.

P.S: If you, oh reader do know AI programming and think my views are hokum (or otherwise), I would love to be told so. I would love comments on this topic, especially as my own knowledge of this field is so inadequate.

Advertisement