Attention is perhaps the most important component of intelligent activity.

There are about 100 trillion synapses and tens of billions of neurons in our brains. The exchange of information occurs, according to neurophysiologists, at approximately 200 Hz in the active mode.

We consider: $ 200 * 100 * 10 ^ {12}=2 * 10 ^ {16 } $ bit. Considering that a neuron is not one arithmetic operation, but a kind of microprocessor, we get that we have petaflops supercomputer "on board". And there is an opinion that exaflops .

No matter how powerful it is, resources are limited. Moreover...

Most of the time, our "supercomputer" is busy with the "calculations" of social interaction.
Speech, hearing, sight, touch, sensations, emotions, empathy...
"Oh! What went! "," Oh! He looked askance at me! ”,“ As you were - you have remained the same. ”
This is neither bad nor good. This is... normal!

Of course, there are deviations, but basically it is.

In order to successfully solve a problem or problem, you need to focus on it. Set clear goals. Stop being distracted, procrastinate.

Or just ignore it (the problem) and do... nothing.

Our brain, in fact, doesn’t care: watch movies, play games or design a spaceship.

ITKarma picture

What goal setting and focus gives:

  • limiting the search area and reducing the combinatorics of possible solutions ( in the elderberry garden, and in Kiev uncle )
  • directing all mental (computational) resources to solving a narrow problem ( whole petaflops! )
  • selection of the optimal instrument, functional basis ( Caesar-Caesarean )
  • filtering, cutting off irrelevant information, although it is useful when there is no solution ( when I eat, I am deaf and silent ).

Goal Selection


"The end justifies the means." The converse is rather false.

There is ambition - it's great! You can dream - great! “But the neighbor.”

But it is impossible to fly to the moon on a bicycle. On foot do not reach America. Bananas do not grow on the apple tree.

Going on a long journey, audit in the garage. Maybe you don’t have a car there. Maybe... the garage itself. As a result, the “long journey” will be a half-hour trip to the store.
The target or target vector is easily described mathematically. This is a list of parameters and their values ​​(specific or range).

For example: height-100mm, width-50mm, weight-0.5kg, speed-3m/s, number of legs - from 3 to 4pcs.
The size of the list may also vary.

The choice of the target vector, of course, depends on the resources available and planned in the future. The goal “Build a roof in 10 days” does not make much sense if there are no walls, but only the foundation. And if there is, please.

To include the parameter "Optimal time of feeding the cat" in the target vector, if you do not have it (the cat) is pointless, exactly the same as the "Amount of water for irrigation of tomatoes" if there is no cottage. It is impossible to make a robot with dimensions of 50x50x50mm if you have only Raspberry Pi as a “brain”.

This is really great! This limits the scope of the search. It focuses on a solution that life already “prompts”, and does not drive into a semblance of schizophrenia “Maybe.”, “And if.”.

Innovation is good - the main thing is not to spend the entire family budget on creating an advanced racing supercar.

ITKarma picture
Setting the threshold of the objective function, which expresses a certain sum - the convolution of a multidimensional target vector.

When the parameters of the target are clearly defined, the process of aiming, pointing or focusing is a matter of technology. Moving forward (the thought process) is carried out by laying a route (searching for neural connections) with holding the target “in the sight”, discarding options that do not lead to the goal.

Also, walking along new “paths”, creating new routes, connections between neurons or their analogues in the circuit is not ruled out.

Stream


Have you noticed when you immerse yourself in the workflow, then you don’t really like when distracted?

This happens because all the “capacities” are occupied, and the process of social interaction decreases in priority. Of course, you can switch, but this takes some time. Plus, to everything else, notorious Flow , which is so hard to achieve.

Thinkers, scientists, programmers generally love solitude - thus, unnecessary information is cut off.
The flow in the information system will be the computation process.
Without power failures, stops, pauses in the algorithm and without communication breaks.
Otherwise, you can interrupt the process at a time close to finding a solution. The results will show that there is no solution, probably the model is wrong and valuable time will be lost.

Finding a solution


Just because a decision comes to our mind does not mean that the brain immediately forms them. enumeration of options of enormous proportions. Often problems are not solved for years and even centuries. Focus + combinatorics. New technologies appear - the functional basis expands.

Many people involved in solving the same problems go to the goal in different ways. If a solution is found, then not everyone. And it’s also true that different people come to similar decisions and even... the same.
If we assume that finding a solution is an absolutely random process and all empirical methods are just ridiculous excuses for scientific “activity”, then a clear goal and focus on it, at least, shows where we are going and the fact of arrival to destination.

image

Time . Also, if we think in terms of goals, then we designate a reasonable time period for their achievements. Usually, this is developed by ourselves by itself, at least for simple tasks like: “Go to the store”, “Hang the chandelier”, “Brush your teeth”.

We don’t do this week, right?
Time is a very convenient parameter for determining such properties as: computing system power, reachability of the target with existing resources, the correctness of the chosen model for calculations.

We can even do such a trick - perform a calculation on 10% of the allocated time on 10 different computing architectures and see, with a high probability, the best system for this task.

Why? Because one architecture may not reach the goal and close.
“This is a failure, Carl!”

Accuracy . It’s probably right to talk about the error of hitting the target. If our target vector is expressed in clear numerical parameters, then we can also evaluate the result.
$ Error=Target - Result $

Percentage accuracy can be defined as: $ Accuracy=100 * (Target Error) )/Target $ . This is an option.

Perfectionism is not a goal achievement . Do not forget about it. This is not about traffic. This is about the accuracy of hitting the target, proximity to the target coordinates. Otherwise, we will spend too much time and other valuable resources, but we will not reach the parameters of the system that we need.

ITKarma picture
By raising the bar high, we risk "turning our neck."

And, as practice shows, the battle for every percent after 80% accuracy complicates the system by an order of magnitude. After 95% accuracy, we are already talking about fractions of a percent.

The reverse side of the focus is the local optimum . Usually we break down a task and solve it in parts. This is a technique. She works. But, achieving the parameters of a separate module that we need, we will not be able to achieve optimal parameters of the system as a whole.
A vivid example: departments of an enterprise. They all have different and rather conflicting goals. And if the director does not competently regulate them, then we will get the result as in the fable "Swan, Pike and Cancer.

ITKarma picture
Local optimum as a result of operation of modules working only for private purposes.

Balance . All of the above concepts must be properly balanced. Everything is good in moderation. Priorities are set for goals that are constantly changing. This is normal.
If you want to live, know how to spin!

If we ourselves do not set goals, then the outside World does it for us.

Many say: “This is karma!” When they constantly fall into the same situations. But if you develop an algorithm of behavior in them, solve the problem, the problem usually "goes away". A new problem appears and the cycle repeats.
Mistakes, frustration - “brain explosion” and urgent restructuring.

In addition, the setting and achievement of goals takes place inside the brain constantly on a small scale. They themselves arise as a result of his activities. Thus, the optimization of the state of the neural network occurs.

Motivation


Our motivation is associated with emotions, which in essence are the total energy of our neural network, the frequency of the brain, sensations that provide a variety of hormones and neurotransmitters. But the external environment, which constantly “kicks us”, is of decisive importance here.

ITKarma picture

The stronger the impact (motive), the more the focus shifts there, the resources are mobilized to solve the problem with the highest priority. An example would be dodging a ball flying at you when you are talking on the phone. The conversation is likely to be interrupted.

Motivation can be internal - a certain abstract problem haunts you. You constantly think about it, look for ways to solve it. Sometimes such thoughts turn into mania, and real needs go by the wayside. Sports fans and rock bands often clatter each other, proving that their idols are the best. Girls don’t like their curlers, guys the size of their... biceps.

The motivation for AI is as follows: we supply an enormous set of data to the input of the ANN and we achieve the maximum of the objective function - the correspondence to the output data obtained and verified as a result of the mental activity of people.

In fact, we give external motivation. We make learning, work the processor and peripherals.

It works! True, this approach does not lead to the creation of something new.
If the result does not suit us, we change the target parameters and (or) retrain. AI remains impassive.
Passions and emotions appear when there is freedom of choice.

The choice of goals and how to achieve them.

Experience is also a source of emotions, it determines the maximum energy, the emotional response to that area (a set of images, a tensor), which constantly deals with biological or silicon brain.

A professional is focused on his field of activity: a doctor treats, an engineer designs devices, a clown entertains people.
Between any areas there is an intersection, general patterns, methods for achieving results.

Everything is interconnected, so we walk in a circle. But if we have a goal, we always know where to go.

That's the whole trick !.

Source