What do you remember most about 2022 in terms oftechnologies? Most likely, many will answer that neural networks. At first, we were delighted with the neural networks that draw by words - probably everyone managed to play enough of the Russian ruDALL-E and other services from our first selection. It turned out that this was not the limit of the ability of neural networks, because then the entire Internet created anime avatars through QQ Small World and stylish portraits using the Lensa application. But it was the Midjourney neural network that made the most noise — in September last year, a man generated a beautiful picture and won an art competition. Neural networks continue to amaze us in 2023. Let's talk a little about how they are superior to people already now?
- 1 The neural network won the photo contest
- 2 The neural network wrote a thesis
- 3 The neural network wrote an essay for admission to the exam
- 4 The ChatGPT neural network was used in court
The neural network won the photo contest
Last year, the image created by the neural networkwon the painting competition. In February 2023, technology went one step further, creating an image so realistic that it won a photo contest.
Every week, an Australian appliance manufacturerDigiDirect is holding a competition for the most beautiful photo with a cash prize. This time, the jury really liked the drone shot of two surfers riding the waves shortly before morning. The author of the picture was identified by a certain Jane Akes.
It turned out that under this name were hidingAbsolutely AI studio staff. After winning, they refused the monetary reward and admitted that the image was created by a neural network. They explained that the beach shown in the picture does not exist, as well as the surfers appearing on it. The authors did not have to get up at dawn, go to the beach, set up the drone and edit the photo. They simply entered a description of the image into a computer program.
The result exceeded all expectations.Not only did the image turn out to be very beautiful, but none of the jury members noticed the catch. This is the first time in history that a neural network has won a photo contest.
According to one of the founders of Absolutely AIJamie Sisson, neural networks cause him fear. He explained that he had won numerous photography competitions and had also won awards for filmmaking. And all his previous work does not look as good as the images generated by neural networks. Today, mankind is at a point where a computer can be a better creator than a person.
And this causes great concern, becauseNeural networks can be used by dishonest people. Absolutely AI has admitted to what they did and refused the prize, but nothing prevents other people from fraudulently winning other contests. Perhaps in the future, the work of people and neural networks will be evaluated separately, in different events. But who can guarantee that among the works of people somehow there will not be images generated by neural networks? Experience shows that often even specialists cannot recognize pictures created by artificial intelligence.
Related article: What dangers of neural networks do we underestimate?
The neural network wrote a thesis
At the end of 2022, OpenAI releasedchat bot ChatGPT, which can give a detailed answer to almost any question. At first, most people were just having fun: asking him to answer the usual questions, generating jokes, and so on. But then neural networks began to find practical applications, such as writing program code and solving school assignments.
In Russia, the student of the Russian State University for the Humanities was the most talked aboutAlexander Zhadan, who decided to use the ChatGPT neural network to write a study on the topic "Analysis and improvement of the management of a gaming company." First, the neural network wrote a diploma plan, and then the student entered queries in English and translated the result - after adding the answers of the neural network together, a document with a uniqueness of 82% was obtained. The head and the diploma commission rated his work as “satisfactory”. After the student revealed the truth, some people demanded that he be stripped of his diploma. But in the end, no one began to cancel the defense of the thesis.
We talked about how to use ChatGPT and what it can do in a separate article. Here is the link.
The neural network wrote an essay for admission to the exam
Also, neural networks already know how to write essays foradmission to the exam. In December 2022, HSE Lyceum teacher Mikhail Pavlovets used the GPT-3 language model (the ChatGPT chatbot is based on GPT-3.5) to write an essay on the topic “Why can progress that gives people convenience and comfort be dangerous for humanity? ". In two minutes, the neural network generated several versions of the text, in which there are links to relevant literary works on the topic. Having chosen the best option, the teacher translated it using Yandex Translator and showed it to his colleagues. Those noted that there are speech errors in the text, but the work deserves credit for all criteria.
ChatGPT neural network used in court
The ChatGPT neural network managed to light up in a judicialdeed. In February, Judge Juan Padilla of the Colombian city of Cartagena asked him a question: “Is a minor diagnosed with autism exempt from paying fees for therapy? The neural network replied that, according to the laws of Colombia, he should not pay for therapy. According to the judge, technologies like ChatGPT can save a lot of time. But the last word must always remain with the judge.
So, in a very short time, neural networks in many wayssurpassed people. Empirically, it was found that in a matter of seconds they can create images that are difficult to distinguish from man-made and win prizes in competitions. Also, neural networks can write essays - it doesn’t turn out perfect, but very quickly, it’s high enough for offset.
You will find more interesting articles on the topic of science and technology in our Zen channel. Have you already subscribed?
What do you think neural networks will be capable of?end of 2023? Can they ever replace artists, writers and programmers? Recently, my colleague Lyubov Sokovikova discussed this topic - I strongly recommend reading it.