The Algorithm of 'Learning'

The mind is a beautiful entity. It perceives music, perceives art, perceives the abstract in a way that perception itself becomes a link between two individuals.The mind is such an entity which enables one to perceive ‘perception’ of another individual long before it has been taught words, taught grammar.The mind does not require words,many times to communicate to another mind. What an elegantly designed system it must be, where an entity is capable of such a thing. Not only that, a mind is capable of ‘reasoning’, ‘judgment’ and also ‘learning’.

What is ‘learning’? Is it the ability to ‘retain’ an experience/information and utilize it strategically in future? Is it the ability to ‘mimic’ another entity, which has been perceived by the mind as ideal? Or is it the ability to combine reasoning and judgment with experience, information and mimicry? In that case, can reasoning and judgment be ‘learnt’?

Most people say, at least I have heard them say..”yes, reasoning & judgment can be learnt…with experience”. The more you practice, more you use your logic, sharper is your ability to reason and better is your judgment, they say. But is that not simply the ability of retention of ‘data’ strategically reused at a different time ..again?

Which brings me to the next question- What exactly is judgment and what exactly is reasoning?

Judgment is like the binary system of 1s and 0s…of good and bad. Classifying an experience, an information, a person as ‘good’ or ‘bad’, just like data being in terms of 1s and 0s is judgment. Just like a certain combination of 1s and 0s gives the final image on the screen, (however oversimplified this may sound) a certain combination of ‘good’s and ‘bad’s enables the mind to reach to the final image of the experience, information or person it is judging. While reasoning is the precursor of judgment (for most humans). ‘Why’ a certain experience/information/person is 1 or 0, ..good or bad.

So considering 1=good and 0=bad, and judgment & reasoning consisting of these two variables, when one says ‘neither good nor bad’ it implies equal numbers of 1s and 0s. ‘Very good’ means relatively higher number of 1s than 0s. ‘Excellent’ being considerably higher numbers of 1s than 0s. Similarly for cases ‘bad’, ‘very bad’ and ‘abysmal’ which brings us to the convenient end result of a rather complex process of ‘judgment’.

Result Process

Neither good nor bad n (0) = n (1)

Good n (0) less than n(1)

Very Good n (0) much less than n(1)

Excellent n (0) much much less than n(1)

Bad n (0) > n (1)

Very Bad n (0) >> n (1)

Abysmal n (0) >>>n (1)

Now, why does a mind classify something as 0 or 1? What is this reasoning?

For me (& I’ve noticed for most people as well), a certain experience/information/person is ‘good’ because it brings ‘happiness’ and a certain experience/information/person is ‘bad’ because it brings us sadness.

Take an example. “Is smoking good or bad?”

Smoking may be ‘bad’ for A and ‘good’ for B.

This is how A’s mind works-

Judgment 1) Smoking causes cancer -> cancer causes pain->pain makes me unhappy. => Smoking =Bad =0

Judgment 2) Smoking causes pollution->pollution causes asthma->asthma makes me unhappy=> Smoking=Bad=0

Judgment 3) I like B. B likes to smoke-> B gets cancer-> B dies-> B’s death makes me unhappy=> Smoking=Bad=0

Judgment4) Smoking looks cool->being cool makes me happy=> Smoking =Good=1

Judgment 5) I have never smoked-> I don’t know what smoking feels like=> Smoking is neither good nor bad.

Here n (0)=3 while n(1)=1 hence n(0)>n(1) which implies final judgment – Smoking =Bad.

While in B’s case the mind might work like this-

Judgment 1) Smoking is cool-> Being cool makes me happy=> Smoking =Good=1

Judgment 2) Smoking feels great->feeling great makes me happy=>Smoking =Good=1

Judgment 3) Smoking causes cancer ->I don’t really care about cancer=> Smoking is neither good nor bad.

Judgment 4) Smoking causes pollution -> I don’t really care about pollution=>Smoking is neither good nor bad.

Judgment 5) I think A is a royal pain in the ass-> I smoke-> I die of cancer-> A wont bug me anymore-> When A does not bug me it makes me happy=> Smoking =Good=1.

Here n (0)=0 while n(1)=3 hence n(0) is less than n(1) which implies final judgment- Smoking =Good.

Now, if you don’t smoke your mind has automatically ‘judged’, using a similar algorithm that A is a nice person=> A=Good while B is an idiot => B=Bad..or if you smoke..your mind may 'judge' otherwise. Why did your mind do that? Even when hypothetical people A & B have nothing to do with your happiness directly. Mind is capable of ‘learning’ to judge & reason from other’s experience/information just because mind statistically foresees what will make us happy or unhappy if we experienced something similar. Not only that, this ‘learning’ is intrinsically biased. Mind relates better automatically to another mind with similar judgment. So if you smoke, you may be relatively happier around other people who smoke. Birds of a feather…..

So it all boils down to ‘happiness’ and ‘unhappiness’ which we call the ‘state of mind’. The ‘state of mind’ after all the algorithms of ‘reason’, ‘judgment’ and ‘learning’ have taken place. It is that what appears on the tiny screen of a calculator when one presses “=” button after a tedious series of calculations. What you’ll do with that figure on the screen, delves into a completely different realm of further ramifications of the 0s & the 1s.

So if ‘good’ (1) or ‘bad’ (0) is happiness based, if reasoning and judgment is “good-bad” based, if learning is reasoning based, the phrase “learning is a lifetime process” must imply that learning is a mere accumulations of these 1s and 0s. So is “learning” a mere accumulations of these 1s and 0s?..and forming a sort of “look-up-table” for all the combinations of 0s and 1s and striving to reproduce instances with combinations where n(0)<<n(1) and striving to avoid instances with combinations where n(0)>n(1)…is this ‘learning’? If it is, then is learning an effort to predict happiness?

We go on “learning” and accumulating these 0s and 1s, these ‘good’s and ‘bad’s , constructing a good-bad binary empire in our minds on a flimsy base material of happiness.

Turns out, we humans are very bad at describing what had made us happy in past, let alone predicting what makes us happy in future.*

So when A makes you happy when you are 10 years old, your 10 year old mind had made a good-bad combination entry in your mind’s look-up-table where n(1)>n(0).

Even after 20 years when A is mentioned in a conversation, considering you have not seen A since you were 10 years old, your mind will go back to the entry dated 9/1/2001, drive: People, Folder: Friends, Object: A and show you the entry A=Good, even if on 9/1/2031 A smokes/gambles/drinks, which according to you may be bad. What will you do then? You will re-enter a new combination of 0s and 1s and modify the entry dated 9/1/2001 drive: People, Folder: Friends, Object: A.

So is this learning? Accepting the fact that we are bad at predicting ‘happiness’ and be willing to change data in our look-up-table?

That brings us to the next level of perception of mind- “Forgiveness” or “selectively ignoring or preventing change in judgment in spite of change in data in look-up-table”.

Now A may drink/gamble/smoke for which different minds will assign different magnitudes of 0s or ‘bad’s i.e for different minds, A’s actions will be judged on a scale of n (0) ranging from a value less than n(1) to a value more than n(1).

Ideally if A’s drinking habits make you unhappy=> n(0)>n (1) => A =Bad

But if A’s drinking habits make you unhappy=>still n(0) less than n(1)=>A=Good.This is an anomaly. An anomaly called ‘Forgiveness’ (sometimes also ‘stupidity’ but most of the time ‘forgiveness’)

Now you may say that A has n(0) less than n(1) for most cases in your look-up-table, hence A should be given a chance to give up drinking habits. This is a second anomaly called ‘Hope’ where mind accepts the first anomaly i.e if A’s drinking habits make you unhappy=> still n(0)less than n(1) => A=Good and projects this into the future making a (statistically unsound) temporary alteration in its look-up-table till A either finally gives up drinking (in which case the data entry will be correct- you are happy-> A=Good & temporary alteration of data has been made permanent) or A continues drinking (in which case the data entry will be incorrect- you are unhappy=> A=Bad drunkard nincompoop)

So is this learning? The ability to take into considerations such anomalies and be willing to alter the state of mind itself as a result of a set of data unlike the previous cases where a set of data resulted from the state of mind.

What causes a state of mind?

How many more ramifications are there to the process of learning? The process that makes us who we are today? What determines its efficiency and what are its limitations.

The perception of perception of mind itself is such a mind-boggling mystery. What finesse of skills must that one have whoever designed this evolving mind of ours.

*This particular line has been borrowed from the talk given by Matthew Taylor. For further referencehttp://www.youtube.com/user/theRSAorg#p/u/1/AC7ANGMy0yo

also the last drawing has been taken from

http://www.toothpastefordinner.com/archives/2004/Aug/

which is quite funny.