Neural networks are taking over the world

The news that a student of the Russian State University for the Humanities wrote a diploma using a neural network and was able to defend it continues to haunt the minds. The capital university proposed to limit the use of the popular chat bot ChatGPT in Russian educational organizations. It was this program that helped Alexander Zhdanov to create the final work on the topic "Analysis and improvement of the management of a gaming company" in 23 hours. Despite the received triple, the young man is not discouraged and claims that his goal was not a good grade, but a desire to test in practice the possibilities of the technology promoted by the American company OpenAI.

Like it or not, and what decision will be made on the controversial work of Zhdanov, the future will show, which from the point of view of the development of neural networks has already come. The defense of a diploma written with the help of artificial intelligence at the Russian State University for the Humanities is not an out of the ordinary case. Sooner or later it had to happen, neural networks are gradually conquering the world and imperceptibly enter our lives. So, for example, the authorities of the Moscow region said that by July they would be able to put AI into the service of tracking stray dogs. Information from CCTV cameras will be accumulated and analyzed for the presence of a leash, muzzle and owner nearby on animals, and then transmitted to catching services. In Colombia, the same ChatGPT was used the other day in court to make a decision on the issue of guardianship. According to Judge Juan Padilla, the conclusion issued by a chatbot on the basis of information about similar precedents processed from the network is not much different from the same conclusion drawn up by people. In any case, the final decision on the case is made by the judge.

Neural networks today already answer not only complex questions and issue conclusions, but also draw pictures (MidJourney, DALL-E), illustrate books, write essays, scripts and even poetry, give medical advice, help programmers write code, etc. and so on. Therefore, the concern of the teaching community, and not only it, becomes understandable. What happens, now machines will be able to perform complex intellectual tasks for a person, and any student will receive a finished thesis from AI in a few hours? Not really. Despite the fact that OpenAI has provided ChatGPT with “memory”, it can maintain a conversation in dialogue mode and make statements based on previously entered information, the chatbot is still far from perfect. For example, according to a study by the Wall Street Journal, he does not know how to joke. What comes out “from under his pen” cannot yet make anyone laugh. In addition, it can give incorrect answers, although they are consistent with all the laws of logic, and their relevance also suffers, because the database must be constantly updated, and the data wired in it still date from last year. Of course, over time, these “misses” will be eliminated one way or another. Therefore, the question of the possible replacement of people by machines, even in complex and previously unconsidered areas of life, remains open.

What happens, in the future we can relax and rely on our kind helpers, who, as worldly wisdom tells us, may one day turn out to be not entirely kind? According to experts, artificial intelligence is unlikely to take away our daily bread and cause ever severe damage. Homo sapiens would not be such if he did not know how to put restrictions on technologies, the development of which along a certain path could be dangerous. For example, restrictions on cloning technology, imposed at one time for ethical reasons, "stopped" its development in a manner dangerous to humanity.

Many people agree that, if used correctly, neural networks will bring undoubted benefits to a person, and of course, they will not be able to exclude him from managing the process. It is in the favorable and enriching "cooperation" of machine and man that our future lies. And now ChatGPT developers, for example, offer teachers and teachers an analogue of "anti-plagiarism" - GPT-2 Output Detector, which can determine whether this or that work was written using neural networks or not. Its accuracy is still lame, but it is still better than a banal ban on the use of neural networks in the learning process.