Automation of science: a robot laboratory assistant and a robot scientist
Homo roboticus
The robot made a discovery for the first time
Irina Yakutenko, <url>Humanity has been afraid that machines will seize power on Earth for a very long time – in fact, since the appearance of the first computers.
A huge number of films and books tell how robots, whose intelligence far exceeds the mental abilities of Homo sapiens, create their own civilizations. People in them, at best, are given the role of slaves. Two papers that appeared on Thursday, April second, in the journal Science, make this prospect much less fantastic.
Science is not a collection of fantastic stories. On the contrary, it is one of the most authoritative scientific journals. The articles published in it are very seriously selected and evaluated by several experts in the relevant fields. That is why the articles of two teams of scientists about the new possibilities of artificial intelligence caused such a resonance in the press.
I came, I saw, I won
The work of a team of authors from the UK is called boring – "Automation of Science" (The Automation of Science). However, its results are not boring. The authors have created, no less, a robot scientist who is able to independently put forward scientific hypotheses, plan the necessary experiments and perform them. In other words, "Adam" (as the designers called their creation) can conduct full-fledged scientific research.
The baptism of fire for the robot was the work on the study of baker's yeast Saccharomyces cerevisiae – a classic object of study of geneticists and molecular biologists. "Adam" had to identify the "parents" of the so-called "orphan" enzymes of S. cerevisiae. This term refers to enzymes for which the genes encoding them have not yet been found. Only by correlating all the enzymes working in the cell with the corresponding genes, scientists can fully simulate the metabolic and biochemical pathways of the body.
The researchers provided Adam with information about yeast metabolism, presented in the form of a list of genes and proteins encoded by them, and also uploaded a database containing a list of genes and proteins involved in the metabolic processes of other organisms. All these data were presented in the logical programming language Prolog, understandable for the robot.
After analyzing the information received, "Adam" became "interested" in 13 proteins and formulated 20 assumptions about which genes can encode them. Next, the robot scientist planned the necessary experiments and performed them.
"Adam" does not look like a "typical" robot, but like a large (very large) laboratory table with many instruments and a computer. Nevertheless, the number of its equipment is limited, respectively, the set of experiments that a robot scientist can conduct is limited. According to the results of the experiments, "Adam" could not unequivocally prove or disprove the hypothesis, but he could estimate its probability. The robot concluded that 8 out of 20 hypotheses can be safely rejected, but the remaining 12 should be looked at more closely.
The creators of "Adam" checked his conclusions by other methods and made sure that their robotic colleague was not mistaken. Moreover, after analyzing the genetic databases that are replenished every day, they found out that six of the "orphan" enzymes studied by "Adam" are no longer orphans. Different groups of scientists have determined which genes are responsible for the synthesis of these proteins. Is it worth saying that "Adam" guessed in all six cases?
The robot scientist has clearly demonstrated that machines can correctly perform laboratory tests without human intervention. Moreover, they do it much faster and with fewer errors. However, even in their article, the authors mention the famous objection of Lady Lovelace, the essence of which boils down to the following: robots cannot generate anything original. (Ada Lovelace lived in the UK in the XIX century. This woman mathematician is known primarily for her description of a computer. In its original form, Lovelace's objection reads as follows: "A computing machine cannot claim to generate anything. She can do all the actions for which people have set the order of their execution.")
In the case of "Adam", we can agree with this argument. Indeed, the robot made relatively simple logical conclusions from the data available to it and developed the most appropriate experiments with the use of available devices. However, it will not be easy for Lady Lovelace to object to the program developed by the second group of scientists.
A machine for understanding the world
Two researchers from Cornell University have created an algorithm that extracts general physical laws from the set of data provided. In their work, the authors tested the ability of the program to comprehend the principles underlying the surrounding world, using the example of a pendulum. Scientists provided the program with data on the movement of components of systems consisting of one or two pendulums (in the latter case, chaotic movements are born). The program tried to describe the behavior of the system using mathematical equations. She had only basic skills at her disposal: she could add, subtract, multiply and divide, but she knew nothing, for example, about theoretical mechanics.
At first, the "explanations" looked almost random: the program tried to describe certain data by selecting more or less suitable mathematical operations. The early equations did not describe the behavior of the system, however, some misses were less unfortunate than others. The program corrected such equations and checked again whether they fit the available data. Gradually, as a result of such selection, she came to such familiar mathematical formulas as Newton's second law or the law of conservation of momentum.
Thus, the program created by scientists, using the simplest mathematical operations, brought out fundamental physical laws from scratch. In this case, we are not talking about any initial logic set by a person. Rather the opposite. To formulate the laws of motion, Newton needed the knowledge gained by generations of scientists who lived before him, and the understanding that everyone who tried to systematize this knowledge was working in the wrong direction.
The program didn't know any of this. She did not have the intuition that helped brilliant scientists to step over traditional ideas about the structure of the world. She just methodically selected possible options and chose the most likely one.
The new algorithm is hardly a universal machine for finding answers to the main questions of the universe. In some cases, it is likely to produce complex formulas that approximate this particular case, but do not reflect the fundamental laws that determine the behavior of the system.
The authors suggest that the program will be useful in those areas where a large amount of experimental data has accumulated, but there are no theories explaining the patterns of their appearance. The scientists said that they applied their development to analyze information about the physiological characteristics of people and the expression level of a large number of human proteins. The results of these experiments have not yet been published, but the researchers said that they were able to discover "some interesting laws", some of which were not known until now.
Time will tell what these two works will mean for humanity. Enthusiasts of artificial intelligence can say that scientists have shown its ability to match the intelligence of humans. Anthropocentrists will object that we are talking only about special cases and it is premature to put an equal sign.
Instead of a conclusion
At the time of it, creatures that did not look like machines lived on Tralfamador. They were unreliable. They were poorly constructed. They were unpredictable. They were short-lived. And these miserable creatures believed that everything must have some purpose and that some goals are higher than others.
These creatures spent almost their entire lives trying to figure out what the purpose of their life was. And every time they found what seemed to them the goal of Life, this goal turned out to be so insignificant and base that the creatures did not know where to go from shame and disgust.
Then, in order not to serve such low purposes, beings began to make machines for these purposes. This gave the creatures the opportunity to serve higher goals at their leisure. But even when they found a higher goal, it still wasn't high enough.
Then they began to make cars for higher purposes. And the machines did everything so accurately that they were eventually entrusted even with the search for the purpose of life of these creatures themselves.
The machines quite honestly gave the answer: in fact, it was not possible to detect any purpose of life in these creatures. Then the creatures began to destroy each other, because they could not come to terms with the purposelessness of their own existence.
They made another discovery: they didn't even really know how to exterminate each other. Then they entrusted this matter to the machines. And the machines are done with this case faster than you can say "Tralfamador".
(Kurt Vonnegut, "Sirens of Titan")
Portal "Eternal youth" www.vechnayamolodost.ru06.04.2009