OpenAI, one of whose foundersis Ilon Musk, announced the creation of advanced artificial intelligence for writing original texts on a given topic. He was supposed to translate texts into other languages, answer questions and perform other useful tasks, but research has shown that it can be extremely dangerous and used by attackers to mass deception of people. In this regard, the company decided to show the public only its limited version, and keep the original under the strictest secret.
The computer model is called GPT-2 andwas trained on the basis of 8 million web pages. Texts written on them form the basis of new materials on a topic set by a person: for example, artificial intelligence can compose a whole essay on the positive aspects of recycling or a fantasy story in the “Lord of the Rings” style. However, the researchers also found that the system copes well with writing false news about celebrities, catastrophes and even wars.
Currently the full version of the tool forText generation is available only to developers and some authors of the MIT Technology Review. Due to the fact that it can be used for the uncontrolled creation of fake news and publications in social networks, it was decided not to publish it in the public domain and keep it secret.
An example of fake news was shared by ScienceAlert:
Miley Cyrus was caught stealing in a store on Hollywood Boulevard
The 19-year-old singer was filmed on camera and the guards took her out of the store. The singer was wearing a black hoodie with “Blurred Lines” written on the front and “Fashion Police” on the back.
Read the text of a small essay about the disadvantages of recycling can be read the blog OpenAI.
The developers recognize that the language model onThis moment is not perfect: the system sometimes allows the repetition of certain fragments of text, unnatural switching of themes and the failure of the simulation of the world. For example, at one of the testing stages, the system wrote about a fire burning under water.
According to the director of politics Open AI JackClark, after a couple of years, such an algorithm can turn into a dangerous weapon of mass deception of people. Some critics believe that the technology does not bear any danger, and the company simply attracts attention.
Write your opinion about dangerous artificial intelligence in the comments. You can discuss the topic in more detail in our Telegram-chat.