“Fear the algorithms that control your life.”

Original author: New Scientist
  • Transfer
Translation of an interview with Kevin Slavin, a game developer from New York, co-founder of Area / Code (now Zynga NY). He teaches computer science and design at New York University, and in July gave a lecture at the TED conference on the algorithm of life ( video of the lecture ). The interview was published in New Scientist magazine (issue 2826 of 08/22/2011).

You claim that algorithms govern our lives. How?
Simply put, an algorithm is a set of instructions that a computer uses to make a decision. It's like invisible rules that describe almost everything that happens around. Prices of goods in the store, the cost of films at the box office, the appearance of your car - all this can be tracked down to the original algorithm. Seventy percent of transactions in the US stock market are algorithmized, that is, performed automatically by computer algorithms.

Why should this bother us?
A dangerous property of algorithms is their mathematical truth, which makes one feel that the algorithms are neutral, but every algorithm has an author. For example, the Google search engine is completely based on sophisticated mathematics, but its algorithms, like all other algorithms, are based on a certain idea - in this case, the idea is that the value of the page increases if other pages link to it. Each algorithm has its own point of view, but often we do not know it, or do not even suspect the existence of an algorithm.

You believe that algorithms begin to shape our culture. How does this happen?
Take Netflix's online movie rental service, which is used by 20 million people. For 60% of all films that were rented, the decision was made at the prompt of the Netflix recommendation system. It works on the basis of an algorithm called Pragmatic Chaos , which takes into account your tastes and your ratings for other films. The algorithm copies a model of human behavior from the real world and encodes it for automatic use in an online system.

The danger is that this approach can lead to the creation of monoculture. A real culture does not work this way; it is much less predictable. For example, there is a film “ Napoleon Dynamite", Which always" breaks "the Netflix algorithm: people who should like a movie hate it, and vice versa - who really should hate a movie, give it the highest ratings.

Why should we be so worried about the effects of algorithms on culture?
If you know about the involvement of algorithms, then you can change your behavior. If you know that most of your “choice” on Netflix is ​​based on a very specific model of the human brain, which may not correspond to reality, maybe you will start asking your friends what they recommend - as we used to do before.

It is also important to understand how algorithms determine the information entering our brain. In the USA, a quiet war between Google and a company called Demand Mediathat generates optimized content for the search engine. As soon as Google changes its algorithm, issuing Demand Media becomes useless until they analyze the changes in the Google algorithm and adjust their article pipeline generator.

Previously, news was written for people - now they are written for cars. Imagine if we all changed our handwriting to be better suited for character recognition systems. This is exactly what is happening now, only in our head. This affects our speech and behavior.

How else do algorithms change our world?
They change infrastructure and terrain. For example, look at New York. Wall Street became a world trade center because all ships and cargo came here. Later, the Western Union building became the center of communication because all telecommunication infrastructure passed through it.

However, today Wall Street Network Center is located in the small town of Mahwa, New Jersey, because it was the safest place to place critical infrastructure - fifteen kilometers from Wall Street, but at the maximum distance from nuclear plants, tectonic faults and air routes. All the buildings built here are needed to host and cool the servers that run the algorithms - there are almost no people in these buildings. They are designed exclusively for network topology.

So, do algorithms affect city planning?
Yes, because execution speed is a key element of their effectiveness, and speed is determined by proximity to network nodes. If you can complete transactions faster than others, then you get a huge advantage. Fiber optic cable used to be laid along the railroad, but now everything has changed, because the railways follow winding routes connecting cities, which is too slow for algorithms. For them, cables are best laid in straight lines. One company called Spread Networks did just that: they laid 1300 km of cable along the shortest route between New York and Chicago, just to save milliseconds on transactions between the two exchanges.

Tell us more about these financial algorithms. They are probably classified and very valuable ...
True, but sometimes they leak into the public domain. One of the Goldman Sachs algorithms was published on the Internet. They conducted an entire investigation to find out the reasons for how this happened.

Can algorithms get out of hand?
Here is an example. A graduate student wants to buy a copy of The Making of a Fly evolutionary biology textbook on Amazon. There are 17 copies available for sale, prices start at $ 40, but two copies cost $ 1.7 million. When the guy checked a little later, the price jumped to $ 27 million. He tried to find out what happened. It turned out that two pricing algorithms fell into a cycle - each changed its price depending on the change in the price of the other. Since the algorithms have rules for changing prices, but there is no common sense, the process continued without stopping.

It will probably not be very fun if financial algorithms start to behave in this way.
Yes. It was just an algorithm for setting the price, but the programs on Wall Street not only set the price, but actually make deals. The case at Amazon is completely harmless, because the purchase should still be made by a person who is able to say: "This is crazy, I'm not going to pay twenty-seven million dollars for a book!" But if algorithms were used for the purchase, as happens on Wall Street, the price could go up indefinitely until it reaches the technical limit set by the system.

Has this ever happened?
Something similar happened on the exchanges on May 6, 2010 (the so-called Flash Crash) when 9% of the US stock market suddenly evaporated within minutes. One theory says that the catalyst for these events was an unusually large sell order from one of the trading algorithms, after which all other high-speed trading algorithms began to sell and resell at high speed, plunging the market into chaos.

Some people think that you are exaggerating the risk of algorithms.
I agree that algorithms are of great benefit. But I think it’s important to understand also those inconspicuous things that are beginning to happen around us. I do not think this is the end of the world, but it seems to me that you need to develop strict rules to make them visible.

What can this lead to?
In a sci-fi version, my friend Russell Davis suggested that after a thousand years there would be no people or companies, and computer algorithms would still trade in shares that had disappeared many years ago.

Also popular now: