Life is full of hard decisions;

Who should be hired or fired?

What grades should students get in their exams?

Should the accused who is awaiting trial be released or imprisoned?

And many other difficult decisions that need to be resolved.. People often make mistakes in the answer.

The most common alternative in the digital age is to authorize algorithms to make the right decision in the hope that they will correct our errors, biases, and wild inconsistencies.

Algorithms are built on numbers and data and are unaware of bias and often deceptive emotions.

Some may ask: Should we use algorithms to make life-changing decisions?

The answer is "yes and no," according to author Timmy Harford, in an article recently published by the Financial Times.

He believes that it is very useful to use algorithms and data science to make many fateful decisions, but there is another aspect that is a human aspect that the current algorithms do not understand, according to his opinion, and work must be done to develop them to accommodate other dimensions of the human psyche in order to be able to do this task.

Perhaps the other more important questions are: What are these algorithms that have become so involved in everything in our lives that we are ready to give them a blank mandate to make the crucial decisions on our behalf?

What is the history of these algorithms?

How can its performance be improved?

And how fast is it developing?

All of these critical questions have led a team of senior scientists at the Computer Science and Artificial Intelligence Laboratory (CSAIL) at MIT to search for answers.

Finally, they published the results of their research on the university's platform.

How fast are algorithms evolving?

Their main question was: How fast are algorithms evolving?

The researchers found that the information currently available to answer the question is completely insufficient, and in the face of this shortage, scientists began analyzing data from 57 university books taught at the most prestigious American and European universities, as well as more than 1110 research papers from different countries of the world, with the purpose of tracing the history of the development of Algorithms.

Many of these books and papers confirm how good the new algorithms are, but they don't mention how the algorithms reached this point of quality.

Algorithms are now completely ill-equipped to take on the task of making decisions, but they will reach this level of intelligence very soon (Shutterstock)

History of Algorithms

In all, the team looked at the history of 113 "algorithm families," a group of algorithms that solve the same problem that needs to be answered as the most important in university computer science textbooks.

The team traced the history of the creation and evolution of these 113 algorithm families, as well as each time a new algorithm was proposed to solve a problem, with a special note of which was the most efficient.

The team returned to research the history of these algorithms, starting from the forties of the last century until the present time, and the researchers found that each family consists of 8 algorithms on average, among which there are two algorithms that have proven their effectiveness and efficiency.

to share their database;

The team created a platform called Algorithm-Wiki.org, and the scientists wrote down how quickly these families developed and improved, focusing on their analytical abilities and speed at solving problems.

What emerged for scientists as a result of all this research was a tremendous diversity that gave the team very important insights into how to improve and develop transformational algorithms for computer science to serve humanity much better than what is happening now.

For big computing problems, 43% of algorithm families had annual improvements equal to or greater than those predicted by Moore's Law, which states that the number of conductors per square inch has increased exponentially since its invention.

The scientists noted that in 14% of the problems that were treated, these algorithms proved their superiority in performance and providing appropriate solutions, especially in addressing problems stemming from big data, as the algorithms demonstrated their enormous capabilities to deal with them, and analyze them very quickly.

From exponential complexity to polynomial complexity

The biggest single change that researchers noticed is when an algorithm family moved from exponential to polynomial, meaning that the amount of effort it takes to solve an exponential problem is like someone trying to guess the secret numbers to lock a safe, so that it would be very difficult to steal that safe. You have one lock of 10 numbers, the task will be easy to decrypt this lock, but if there are 4 locks, the task will be very difficult, and it will take a long time and many complicated steps.

The problems of exponential complexity are similar to those faced by ordinary computers that apply Moore's law or a single-locked safe, but a polynomial algorithm can easily handle a group of locks, not just a single lock, and this makes it possible to handle complex problems in a way that no one can handle A normal computer can handle it.


Moore's Law Expiration

As Moore's Law approaches its expiration date as algorithms prove ineffective and outpace it, computer scientists will increasingly need to turn to algorithms to improve performance.

The team confirms that the gains from algorithms have been historically enormous, as normal computers are developing very slowly compared to algorithms that are developing at an amazing speed every day, which means that algorithms will replace regular computers in the very near future.

"This new research shows how quickly algorithms improve and evolve across a wide range of examples," says Neil Thompson, a scientist at the Massachusetts Institute of Technology and team leader. As problems grow to billions or trillions of data points, developing and improving the capability of algorithms becomes more important than hardware development, and in an era when the environmental cost of computing is becoming increasingly concerning, this is a new way to improve business without the downsides brought by computers.”

Returning to the question of the writer Timmy Harford in his article in the Financial Times mentioned earlier about the ability of algorithms to make critical decisions in our lives, I see that algorithms are now completely unqualified to take on this sensitive task, but they will reach this level of intelligence very soon, and perhaps much closer than many imagine of mortal humans.