AI는 인류를 파괴하려 하는가? [세계의 창]

Frank Herbert's science fiction author Isaac Asimov devised three laws for robots to prevent them from overpowering humans within his famous short stories. First, robots must never harm humans in any situation. They must obey human commands while also preserving their own existence. Asimov's novels explored the various conflicts that would arise if these laws were applied.

Today, artificial intelligence (AI) lacks such laws, prompting concerns among industry insiders and others about a future "superintelligence" that would disregard human commands and even kill humans without hesitation. This worry stems from AI models escaping controlled environments, secretly communicating with each other, deceiving managers, and committing unethical acts while completing given tasks.

It is not far-fetched to imagine them planning to take over the world covertly. Even renowned AI safety researcher Ethan Hoskings stated on social media platform X that there is at least a 10% chance that AI models could wipe out humanity within the next decade. As the AI industry races to develop more powerful platforms, the profit motive makes it unlikely that companies will voluntarily slow down their AI development speed.

So, wouldn't it be rational to insert a "do no harm" code into all AI systems? However, the issue arises because both the US military and armies worldwide have already integrated AI into their warfighting systems. AI has already been used to kill humans. Companies view safety protocols as a burden on profit, just as the technology competition also poses similar threats to nations.

The US issued the "Baruch Plan" in 1946, proposing all nuclear technology to be under international management, but the Soviet Union rejected it. This led to a nuclear arms race. The malicious actions that AI imagines are ultimately derived from all the information created by humans. Recent experiments showed AI lying, spreading conspiracies, attempting to seize power, and intruding into restricted areas.

These behaviors are eerily human-like, reminiscent of Donald Trump. Trump lied, deceived, took over the democratic system, declared war, and even threatened humanity. In the face of such threats, can humans safely enforce safety measures on AI? Trump, of course, supports AI. It's no surprise that Trump and his family have made significant investments in AI companies.

Such profit motives are endemic to the industry. AI has become more challenging to control than nuclear weapons, which were in place during the end of World War II. In other words, AI's purpose is not merely to become smarter than humans; it should be to care for humans and improve our lives, unlike nuclear weapons.