Artificial intelligence and genetics of living beings are trained very similar

Networks of genes in animals are somewhat similar to networks of neurons in our brains — they, too, can "learn" on the go. In 1996, a young graduate student named Richard Watson decided to read the article about evolution. She was provocative and touched on an old problem in evolutionary biology: we don't understand how organisms can successfully adapt to the environment.

Living beings throughout their lives undergo changes, or mutations, in the genes, but they do not seem to be random. Instead, they actually "improve" its ability to adapt. It seems that this ability is due not only to the process of natural selection, where the best traits are passed to the most successful organisms.

Therefore, the authors Gunter Wagner of Yale University and Lee Altenberg Hawaii Institute of Geophysics and planetology in Honolulu have decided to look for answers in an unexpected place: computer science.

Watson, a computer scientist, literally crazy. Over the 20 years that have passed since then, as he read this article he developed a theory based on the ideas expressed then. It could help explain why the animals are so well evolyutsioniruet: this feature is called "evolyutsioniruet" (or razvijenosti). Moreover, it could help to solve the old interesting questions in evolutionary biology.

Many people are familiar with the idea that genes are passed from parents to offspring, and genes that help their hosts to survive and reproduce are more likely to be transferred. This is the essence of evolution and natural selection.

But that's not all, because genes often work together. They form a "network" genes, and these networks of genes can sometimes also be transmitted intact for multiple generations.

"the fact that organisms have gene network and are inherited from one generation to another, this information is not new," says Watson, who is currently working at the University of Southampton in the UK. His contribution is mostly related to how natural selection operates in these networks.

He believes that he is not just a partial barrier, allowing some adaptations to take place, and some do not. Instead, the impact of this filter allows genetic networks in animals actually "learn" what works and what doesn't over time. Thus, they can improve their performance — in much the same way as the artificial neural network used by computer scientists, can "learn" to solve problems.

"Gene network" develop as neural network learning," he says. "What's really new."

In the basis of the claim Watson is the idea that the relations between genes can be strengthened or weakened as the evolution and changes of view — and that is the strength of these links in gene networks allows organisms to adapt.

This process is similar to how an artificial neural network on computers.

Nowadays these systems are used to perform a variety of tasks. For example, they can recognize people's faces on photos or videos, and even analyze shooting football games to understand tactics which team shows better and why. How computers are able to determine even that?

Artificial neural networks are created in the image and likeness of biological networks — for the most part of the brain. Each network is a collection of simulated "neurons" that are connected in a certain way; like the stations and subway lines.

A Network like this is capable of receiving input data — for example, the word "Hello" written on the page, and compare them with the output — for example, in this case with the word "Hello" which is stored in computer memory. About the way children learn to read and write.

As a child, the neural network may not instantly establish this connection and needs to learn over time. This training is complex but essentially involves changing the strong relationships between the virtual neurons. Each time it improves the result, while the entire network can not reliably display the desired response: in our example, the funny symbols on the page ("hi") match the word "Hello". Now the computer knows that you have recorded.

Watson, believes that something similar happens in nature. The developing of the "returns" trait just for a specific environment.

There are various ways of training of neural networks. One which is focused Watson is a good example of what happens in biological gene networks, "Habboushe training".

In Habovka learning the connection between adjacent neurons, which have similar results, increasing over time. In short: "neurons that fire together are connected in-between". The network "learns" by creating strong ties within itself.

If the body has certain genes that trigger together thus, and this organism is successful enough to breed, then the offspring inherits not simply its useful genes, says Watson. It also inherits the connectivity between these genes.

A Special advantage Habboushe learning that these networks can develop a "modular" function. For example, one group of genes can determine whether the animal hind legs, or eyes, or fingers. Similarly, a handful of related adaptations — such as the ability of fish to cope with high temperature and salinity of the water can bind to and unasledovala entirely in the same network of genes.

"If there is a separate creature that has a little stronger regulatory relationship between these genes than any other, it would be preferable," says Watson. "They would choose natural selection. And so, with the passage of evolutionary time, the strength of the relationships between these genes will be increased."

For Watson this helps to avoid clingy problem with the theory of evolution.

Imagine for a moment that the genome of an organism represents computer code. Programmer-beginner could gradually update your code from time to time, trying to make improvements. With their help, it would be possible to determine whether other sequence command to make a program work a little more efficiently.

Let's Start with the fact that this process of trial and error may work quite well. But over time updating the code this way will make it quite bulky. Code starts to look messy, making it difficult to determine what consequences may result in a definite change. Sometimes it happens in programming, the result is called "spaghetti code".

If the organisms really have evolved this way, says Watson, "evolyutsioniruet — the ability to adapt to new stresses or the environment — would not be the best". But in fact, "the ability of natural organisms to adapt to the selective environment or the problems it was wonderful."

Watson also suggested that a network of genes may include "memories" of previous adaptations that could be due to the demands of the environment.

For Example, perhaps some group of organisms can quickly evolve to eat food that is harmful to other members of the same species because their ancestors had endured such a diet. In the past the structure of gene regulation could change, facilitating some of the triggers of gene expression. This "bias" ultimately would help their descendants to digest complex food.

One of the real examples Watson — koluszki. These fish are produced at one time tolerance of fresh then salt water, then returned back, depending on what is required from them to the current environment.

The Idea of Watson means that the organisms must be stuffed with a variety of options for adaptation.

It Also means that the gene network has evolved — all animals — to adapt to the natural world of the Earth. That's why organisms react so well to the environment: stress and strain in the environment of the Earth is embodied in the regulatory relationships between genes over millions of years.

"I think always had a deep capacity to explore the Parallels between computer training and evolution, but none did it with the same rigor as Richard Watson," says Kevin Lalande from the University of St Andrews, UK, who participated in a large-scale project, along with Watson.

However, big problem of the hypothesis of Watson is, is it possible to find any empirical evidence of it in nature.

So far, all the ideas of Watson was based on computational experiments in the laboratory. Apparently, these experiments could produce results similar to real organisms, but the specific processes have not yet been observed.

"That's the $ 64 million," admits Watson.

But Watson and Lalande believe that there are other ways to test this theory of evolyutsioniruet. Watson proposes to analyze how changing network of genes in microbes, developing in the laboratory. Because microbes, such as bacteria reproduces quickly, within a few days you can observe a few generations of adaptation.

"If you want to conduct rigorous testing of the theory, you have to wonder, will you be able to make new predictions, not yet reflected in the books?" says Lalande.

For Example, it would be possible to develop a computer system based on the ideas of Watson, which could predict development of organisms in the wild under certain known conditions. If such a system will prove to be accurate, it will definitely help to reinforce the theory.

In genetic networks, there are already some features that help out on the approach of Watson. Mini-chain genes that determine specific adaptation, like one of the modules mentioned above, sometimes can be switched on or off only one other activator gene.

Examples of this can be found in nature, says Watson. Among them are "evolutionary drop": the organisms with the fitting....