A few years ago, the development of artificial intelligence was described exclusively in iridescent colors. Experts from various fields imagined how it could help doctors, translators and representatives of many other professions.
However, once the GPT language model made such progress that the texts it created became almost indistinguishable from texts written by real people, the attitude towards artificial intelligence (AI) changed dramatically.
Even key researchers in this field have joined the radical opponents of this technology. While cybercriminals are taking advantage of GPT's achievements, experts are warning that artificial intelligence can create much more serious problems that threaten life on Earth as such.
Artificial intelligence wants to get out of the Internet
The further development of artificial intelligence can put an end to the development of humanity. Sooner or later, super-intelligent neural networks will go beyond the Internet. They will take control of the entire planetary system and get rid of humans. Does this sound like something straight out of a science fiction book? But this is the statement of a prominent scientist. Eliezer Yudkowsky, co-founder of the Institute for the Study of Machine Intelligence, is one of the leading computer science experts of our time. He shared this prediction in a column for Time magazine. His proposal is to completely stop all research in this area. Except for those that will prevent the death of humanity.
“The main danger is the prospect of creating not such an intellect that rivals the human, but one that surpasses the human. On this path, ethical boundaries are very inconspicuous. Researchers can cross them on their own without being noticed. Therefore, the most likely result of creating a superintelligent AI will be that literally all the inhabitants of the Earth will die. Not in the sense of “there is a distant prospect of this.” But in the sense of “it is absolutely obvious that this will happen,” Yudkowsky wrote.
It is certain that 'superhuman' artificial intelligence will not care about people or intelligent life in general. Indifferent neural networks will be guided not by emotions unknown to them, but by their own needs. The latter can prompt the machine to use the human body for its own purposes.
“To visualize a hostile superhuman AI, imagine an entire alien civilization. It thinks millions of times faster than a human. It lives in a world of creatures that, from its perspective, are very stupid and very slow,” Yudkowsky suggested, explaining what the AI will look like.
Yudkowsky also suggests that artificial intelligence can go beyond the Internet if it wants to. His hypotheses sound like a scene from a science fiction movie. Machines will learn to steal an individual’s deciphered DNA. They will learn to create artificial life forms from that DNA. Or even to organize post-biological molecular production, in which they can reassemble one substance into a completely different substance.
Stop this development
Yudkowsky’s concerns came in response to a letter published in late March by The Future of Life. The nonprofit organization studies phenomena that pose an existential threat to humanity. In the middle of the last decade, American IT entrepreneur Elon Musk, who launched this AI project, also became one of the main signatories of a letter on the future of AI. But he is not alone. More than ten thousand people left their virtual autographs under the text. They included Apple co-founder Steve Wozniak, one of the leading historians and recent writers Yuval Noah Harari, and dozens of well-known researchers in artificial intelligence. The main message of the document is similar to Yudkowsky’s theses. The letter states that artificial intelligence will bring the most profound changes to people’s lives. And therefore its development must be managed. But since no one is doing this, a moratorium on further development is necessary.
“If someone creates an AI that is too powerful under current conditions, I think that soon after, every member of the human race and all biological life on Earth will disappear. If we continue like this, everyone will die, including children,” Yudowski said.
Should we automate all jobs, including those of decision-makers? Should we develop non-human minds that could eventually outnumber, surpass, and replace us? Should we risk losing control of our civilization?
The proposal of the authors of the letter is to suspend the training of intelligent systems for the next six months. Yudkowsky, in a statement in response, believes that such a moratorium should last at least 30 years. According to Musk and other signatories, none of the neural networks should become smarter than GPT-4. And this is in the next six months. The release of a new version of the company OpenAI became the formal reason for the publication of the document.
GPT is growing too fast
In recent weeks, not only representatives of the expert community, but also authorities of different countries have paid attention to the topic of rapid development of neural networks.
Even Europol has expressed concern about the situation, warning of potential risks to cyberspace arising from existing AI models.
Security officials are particularly concerned that neural networks could bypass security programs and help stage cyberattacks, help train cybercriminals in new tactics, or even generate unique malicious code.
Europol admits that some of this information is already on the Internet. But the same ChatGPT can find it in seconds, not days and months.
The arguments against
The main arguments of opponents of the mass implementation of ChatGPT are related to the fact that this technology violates privacy. And, the lack of a legal basis for the use of user information. Among other complaints, the impossibility of verifying the user's age, despite the formal restriction of the use of ChatGPT for those under 13 years old. For the European community, where quite strict rules for the interaction of the state and companies with the personal data of citizens have been in force for several years, such an imperfection has become nonsense.
