Prediction Machines
A look into why a lot gets unsaid at the end.
Unspoken expectations are premeditated resentments.
Neil Strauss
You gather all courage. Did the mirror talk. Conversed with the ancestors and all the animals infesting your lovely home. Paced 200 times. Made sure that this was finally the right time. The stars have not aligned but my ooh my.
You can no longer hold this on your own. You need to tell someone. I mean, you have told every non-living being. The showers can recite your worries. The doors sing of your frustrations. The windows whispers sounds of comfort( but they be just whispers(you do not give a damn))
It has eaten you alive. I mean you’ve faced the devil and all his noisy neighbors. Surely, this would be the least of your concerns. You see them come. It takes what seems like eternity. I mean they took off from earth and landed in mars before they sat down. You force a smile for the sake of who knows what. Then cold starts creeping. But the day is as hot as hell. The utterance begins. Your thought has second thoughts. The words hide. The stomach is a safer place for them. So you take a breath, your tongue embraces both your lips, you take a sip of whatever is close.
Then.
Finally. Fiiiinnnnaaaaly
You start. So… eeh… Where do I begin…
You make some adjustments. Fidgeting becomes second nature. You wonder what’s wrong with the chair. The table keeps moving. The world is finally revolving.
One eternity later. After eons of rambles, shambles, yappings and overexplanations, you decide your words carry no meaning no longer.
You allow both lips to have a reunion. You wait.
The next thing you know, you are in the Afghanistan war. Bullets and grenades everywhere. Missiles and sounds of choppers in the brutal night. Chaos. Utter chaos! It’s attack after attack.
Well, you are utterly baffled.
You were just…
The reason why you get fed up trying to make another understand your perspective, is based on re-enforced learning at the very least.
We forget that machines are modeled to mirror humans. We have that intrinsic property from which they get modeled.
Inputs and outputs. For some input, we expect a certain output.
Inputs & Outputs
Therefore, given that input X was a detailed explanation of why you did what you did(reasons) and expected output Y(expected feedback) was at least an “understood”, if given output is negative(not what was expected), the cognitive system starts learning.
What does it learn?
This is somewhat complex.
It varies.
Fundamentals
But on a more fundamental level, the system learns that input A is either good or bad. Or neutral.
In a mathematical representation; 1, 0 and -1.
I do not seek to reduce humans to just simple models, however, this seems to show up again and again.
It’s a system that sort of works.
We survive because we learn who to trust and when. Those who did not, went extinct.
Trust comes up as very fundamental.
Take care of those to our right and left. When we take care of the group, the group takes care of us.
Simon Sinek
Why does it learn?
The more important question is, what happens when it does not?
Repetition. Repetition of given behavior.
If a child does not change input(behavior), after several burns from a fire or cuts from a blade, the child’s ability to perform certain tasks will be taken away from him/her.
Another possibility is death. When severity comes into play.
When does it learn?
We’ve kinda touched on this. It is when what was expected is not fulfilled. When the thirst is not quenched. When a promise is not resolved! We realize that we need to update the model of the world.(Sometimes we learn the wrong stuff)
How does it learn?
Humans are prediction machines with capacity for adjustments given new evidence.
This is where GIGO(garbage in garbage out) comes into play. When any model trains on “bad” data, you can try to increase the amount of “good” data it receives. Of course, good and bad may be objective or subjective based on the type of system in play.
Anyway, the most recommended way of dealing with this is by throwing away the bad system. It comes with the whole sunk cost fallacy all over. However, what are we optimizing for? Suffering in the short term for the sake of a proper path seems like a pretty good deal. (Again, subjective) So “new clean data” is used to train this system. Again nothing is guaranteed.
I am fully aware that this is not a luxury we as human beings have; to just throw away all the learnt experiences we had. However, we can unlearn, relearn and learn new stuff. We are beasts of nature. Susceptible to the amygdala, yes, but potential exists. Idk what the clear answer is. Moving in the direction of that which maximizes success is a pretty good heuristic.
Whatever doesn’t kill you makes you stronger
Someone
The worst thing that has ever happened to you has happened to you.
Sometimes we over index on the negative and romanticize the past. I know. Where do we draw the line between reasons and excuses(for our and other people’s actions)
Back to machines. There are things called weights. Yes, it probably works the way you think it does. For every input S, there exists a corresponding weight W. Things with high correlation get higher weights. For example, in furniture, tables and chairs will have higher weights compared to man, maize or ocean.
And that’s how change in behavior begins.
The sad thing is, this negative feedback if repeated enough times, no longer just applies to person X, it gets set in stone as another rule to be followed religiously.
The bites
Remember, once bitten twice shy.
I believe that once it gets incremented, the effect is not linear, ie, twice maps to thrice.
It is more likely to be a X^n, where n is the number of repetitions and X is a number north of 2 (and more likely to be a 10).
And trust falls logarithmically. That is with every addition, we move sharply down to zero.
Obviously, this has no scientific or mathematical evidence.
This ties well with: fool me once shame on you, fool me twice, shame on me. Cause now I’ve kinda become the kind of person who gets fooled/lied to etc. That’s why telling someone “trust me” after breaking their trust is met with for lack of a better word - confusion. Wdym, trust you! You might as well pull the trigger at this point. (Learnt this. It sucks)
Risk reward
Based on evolution, we are tuned to risk reward tendencies.
For given risk A, we got reward B. The reward could be positive or negative, to the extent of death. If reward is good, repeat, if bad, avoid. Repeat.
Classic example;
Cats are afraid of cucumbers. This is because they confuse them with snakes. The generation of cats that had false negatives, that is, it’s just a cucumber and ended up being a snake, ended up dead, and did not live long enough to reproduce. So, we are left with an anxious generation that uses the logic: if it looks like a snake, it is a snake.
Another I heard from J. Malen.
If it walks like a duck, quacks like a duck, eats like a duck, then it is a duck.
If someone shows you who they are, believe them.
Someone
Survival
Evolution performs a loose fitting function. You just need to survive long enough to reproduce in order for your species to survive.
Rory Sutherland
The same sort of applies to humans. A child and even adults at any age, have similar responses. Given X, I expect Y. If not Y, adjust to fit output.
The best thing is to reward good behavior and punish bad ones.
Issue is, who defines good or bad?
Hold on
Obviously, we are not the saints in most cases, consequently, the treat others as you’d want to be treated applies.
Or the Jesus quote,
Love thy neighbor as thyself.
Jesus
(I sometimes wonder if this rule would apply for someone who hates themselves and aims at causing harm to themselves. Surely, they ought to stop at love thy neighbor, or at least, they need really kind and loving neighbors)
Ok. What now?
This takes us to this fundamental question: Who/what defines us?
In systems, the maker of a thing knows the purpose of a thing and how it works. Sometimes, there are surprises(mostly bugs and learnings)
Anyway, there exists internal and external conditions.
Internal conditions:
Things within our control eg getting outside, people we interact with etc
External conditions:
Things outside our control eg the weather, people’s words and reactions effect