According to Sam Altman, government regulators should prevent dangerous consequences for humanity from such language patterns. ChatGPT answers 90% of the bar exam questions correctly and is virtually error-free on the high school math test.
Sam Altman, CEO of OpenAI, which developed the ChatGPT AI-based application, warned of the dangers of technologies that are changing society. It is reported by ABC News.
Altman, 37, stressed that government regulators and society must work together to prevent potentially dangerous consequences for humanity.
“We have to be careful with this. <…> I am particularly concerned that such models can be used for large-scale disinformation, Altman said. “Now they write better computer code and can be used in cyberattacks.”
Despite all the risks, ChatGPT could be the “greatest technology” that mankind has managed to develop so far, he said.
Interview with Sam Altman
Such a warning from one of the authors of this technology came after OpenAI introduced the latest version of the GPT-4 AI language model, less than four months after the release of its first version, which became the fastest growing consumer application in history.
In an interview, an OpenAI representative noted that the development is not perfect, but at the same time, during testing, it answered correctly 90% of the questions for the bar exam. And she practically did not make mistakes during the school test in mathematics. Moreover, AI can write codes in almost all programming languages.
Most of the fears associated with the development of the field of artificial intelligence boil down to the fact that AI will completely replace humans. However, Altman assures that the system is completely under human control.
“AI is waiting for someone to give input. It is a tool that is controlled by a person,” he explained, adding that the whole question is who controls it.
“There will be other people who do not adhere to the restrictions that we adhere to. I think society does not have enough time to understand how to deal with this, how to regulate it,” he explained.
Recall that the ChatGPT chatbot passed the captcha, claiming to be blind and hiring a freelancer. For the first time, AI offered to help a real person to circumvent Internet security measures.
Focus previously reported that Microsoft has built ChatGPT-4 AI into office products. In Microsoft 365 Copilot, the neural network is used for all Office tasks.
