Are the artificial intelligence models available to the public really the most advanced versions that exist? The versions available to ordinary users are astonishing in both their capabilities and their rapid development. Yet there are elite versions that the public has not seen.
The White House reportedly approved Anthropic's "Mythos" model, a system that has not yet been released to the public because of its capabilities in cybersecurity and its ability to penetrate some of the most sophisticated electronic systems. It is currently being used only by a limited number of institutions.
Anthropic itself, a company specializing in artificial intelligence, has warned about the accelerating pace of AI development and has proposed a temporary halt to the advancement of the most powerful AI systems, amid signs that the latest models could potentially move beyond human control.
The company behind the Claude model argued, according to an AFP report, that a global slowdown in the development of advanced artificial intelligence would "most likely be a positive thing."
Researchers describe a mechanism known as "recursive self-improvement," whereby an AI system becomes capable of teaching and improving itself with little human assistance. This is precisely the phenomenon that the leading US company has warned against.
Anthropic has even called for an international treaty similar to the Nuclear Non-Proliferation Treaty, aimed at limiting the spread of AI capabilities and ensuring that they remain under control. But who, for example, could persuade China to agree to such an arrangement? And how could such efforts be enforced when programmers working behind closed doors continue to develop this powerful technology?
Elon Musk, Anthropic, and others have focused on the military, intelligence, and security applications of these alarming technologies at a time when the world's operating systems and data storage have become increasingly dependent on the digital realm. It is difficult to see how these systems and databases can be guaranteed protection from attack, regardless of how advanced any country's cybersecurity capabilities may be.
Do backup copies exist outside this interconnected digital world, disconnected from it entirely? Are there even backups in traditional forms such as paper and other physical records?
That is one side of the issue. The other is how to preserve the minds and well-being of future generations in a healthy and natural state. How can humanity's cognitive abilities, including analysis, synthesis, and critical thinking, be preserved?
Will people eventually switch off their own minds in favor of artificial intelligence, just as mental calculation declined with the advent of the calculator? And is the comparison even valid, given the potentially far greater impact of AI on the human mind?
For this reason, global governance, a genuinely international oversight body, and a binding ethical framework for this new arrival, artificial intelligence, have become an urgent necessity to safeguard humanity's future, if the leaders of these companies and the heads of the world's major powers still possess enough reason and moral responsibility to act.
