Calls on Elon Musk to stop developing “dangerous” artificial intelligence

AI systems with human competitive intelligence can pose serious risks to society and humanity, as evidenced by extensive research (1) and recognized by top AI laboratories.

As stated in the widely adopted Asilomar AI Principles, advanced AI can represent a profound change in the history of life on Earth, and must be planned and managed with due care and resources. Unfortunately, this level of planning and management does not happen, although recent months have seen AI Labs race out of control to develop and deploy more powerful digital brains that no one – not even their creators – can reliably understand, predict or control.

Contemporary AI systems are now able to compete with humans in public tasks, and we have to ask ourselves: Should we allow machines to flood our information channels with propaganda and lies? Should we automate all jobs, including the ones that do? Should we develop non-human minds that may eventually outnumber, surpass intelligence, outlive, and replace us? Should we risk losing control of our civilization? Such decisions should not be delegated to unelected technical leaders. Strong AI systems should only be developed once we are confident that their effects will be positive and that their risks will be under control. This confidence must be well justified and increase with the magnitude of the system’s potential effects. OpenAI’s recent statement on AI states that “At some point, it may be important to obtain an independent review before starting to train future systems, and for more advanced efforts to agree to limit the growth rate of computing used to create models.” We agree. The point now.

Therefore, we call on all AI labs to immediately stop training AI systems more powerful than GPT-4 for at least 6 months. This discontinuation must be public and verifiable, and include all key actors. If such a moratorium cannot be triggered quickly, governments must step in and impose a moratorium.

AI Labs and independent experts shall use this pause to develop and implement a set of joint safety protocols for the design and development of advanced AI that are rigorously audited and supervised by independent external experts. These protocols must ensure that the systems they adhere to are secure beyond a reasonable doubt.

This does not mean pausing the development of artificial intelligence in general, but rather simply stepping back from the dangerous race to larger and unpredictable black box models with emerging capabilities.

AI research and development must be refocused on making today’s robust, modern systems more accurate, secure, interpretable, transparent, robust, consistent, trustworthy, and loyal.

In parallel, AI developers should work with policymakers to accelerate the development of strong AI governance systems. These should include at a minimum the following: new and capable AI-dedicated regulatory authorities; Oversee and track highly capable AI systems and large sets of computational abilities; source systems and watermarks to help distinguish real from synthetic and to track model leaks; a robust auditing and certification ecosystem; liability for harm caused by artificial intelligence; strong public funding for technical safety research for AI; and well-resourced institutions to deal with the dramatic economic and political (particularly democracy) disruptions that AI will cause.

Humanity can enjoy a prosperous future with artificial intelligence. Having successfully created robust AI systems, we can now enjoy an “AI summer” in which we reap the rewards, engineer these systems to achieve the obvious benefit of all, and give society a chance to adapt. Society has temporarily stopped using other technologies with potentially catastrophic effects on society.

We can do that here. Let’s enjoy the long summer of artificial intelligence, not rush unprepared for fall.