Bill Gates says AI is worse than nuclear weapons

Bill Gates, co-founder of Microsoft and chairman of the Gates Foundation, in an interview with The Verge called for preparing for about 20 years of turbulence caused by the development of artificial intelligence.

In an essay in late August, he wrote that even in the best-case scenario, the transition to a new era of AI would be one of the most tumultuous periods in human history. In subsequent conversations, he said that humanity has never met anything more dangerous than AI, and noted that even nuclear weapons against this background look less serious threat.

He said progress over the past year has been astonishing. AI has learned to write code, look for cyber vulnerabilities, reason in biology and build relationships with people. Gates says the current incredible power means it will all happen and society must be ready. The main risk, in his opinion, comes from models without protective mechanisms, which can remove all the restraints. They pave the way for cyberattacks and bioterrorism. The more distant threat is the loss of control over any advanced model.

Gates stressed that humanity has overcome the limit of possibilities. He sees nothing more dangerous than all human-made technologies and says that AI will affect all areas of the economy and change the nature of wars. He said the industry is open about risks, but outside the industry, the willingness to believe that this time is different is very low. He added that the warnings from industry insiders are not insincere and are absolutely true.

To control AI, Gates identifies three fundamental approaches. You can change the way you build models and move away from reinforcement learning. You can provide transparency and make the model a white box to see what's inside. You can create a “police layer” of symbolic security rules that the model can’t break. The latter option, he believes, is the most likely solution to the leveling problem. At the same time, Gates admits that today we are not ready for alignment, because we are not good enough and there is a lot of work to be done.

To reduce risks, he proposes to introduce mandatory controls for models above a certain level of capabilities. It will be possible to run them only on platforms that certify the preservation of protective mechanisms even during customization. When asked about the nihilism of the younger generation, Gates admitted that the next 20 years will be a serious shock to society. Even with better AI management, this will be one of the most tumultuous periods in history. He advises young people to apply AI where it cannot yet replace them, and to influence politics.

OpenAI model hacked test wallet to pay commission

Reliable news is always at hand - in the channel "MK" in MAX