من هم الناجون من خطر الذكاء الاصطناعي؟

Why are calls to slow down artificial intelligence increasing? Are there countries, companies, or individuals that are safe from its dangers? Why have warnings and calls to slow down AI and the pace of developing more advanced models come from some of those who create and lead its largest laboratories, rather than from technology opponents and critics?

I pose these questions after the CEO of Anthropic called for independent evaluators to be introduced into AI companies and for enhanced coordination between companies and nations. His call has received support from other leaders in the AI industry, while political and technical voices have opposed it, fearing that slowing down could lead to a loss in the technological and techno-political race. This very division reveals the essence of the problem.

There is a fear of advancing too quickly, which causes risks, as well as a fear of falling behind in competition among tech companies and nations that are betting on AI for superiority and dominance amid the fierce competition the world is witnessing.

The capabilities of AI have been advancing much faster than the progress of legislation, testing, safety standards, and international agreements. So, could it be that the goal of the calls to slow down is to buy time and establish international rules before AI enters a cognitive and technological arms race from which no one can withdraw alone?

In a global race of this magnitude, the motives of safety intertwine with those of competition, economics with national security, and regulation with attempts to protect market positions. We are not facing a disaster that has already occurred, as some open sources claim, but rather a highly impactful possibility, with unknown probabilities and timing. Herein lies the danger and difficulty. The fear may lie in what we voluntarily concede in our decisions to AI, alongside relying on recommendations, analyses, or predictions without human capacity for understanding, reviewing, questioning, and verifying—essentially, an unconditional reliance on AI.

If a highly capable AI system goes out of control, who are the candidates for survival? Who will be safe from danger? Are those less connected to technology safer? Are advanced countries safer than developing ones? Are tech companies safer than others, or are it the communities that use technology less?

The answer: Unfortunately, there is no category that can be guaranteed safety and security from danger automatically, according to the data! However, there is a difference between direct exposure and indirect exposure.

A person who does not rely on AI in their work may be less exposed to some of its direct errors. An institution that does not grant intelligent systems access to its funds or operational infrastructure is less exposed to direct failure. But this is relative safety and very relative protection; because a person who does not use AI lives in an economy that banks use and buys from companies that also use it, lives in a country that uses digital networks, is affected by financial markets that use algorithms, and consumes media that AI may produce part of. They rely on communications, energy, transport, and global supply chains that are highly interconnected. Therefore, you may choose not to use AI, but it is difficult to live and work in a world that does not use it.

The strange paradox is that the most advanced may be more exposed to danger. A highly digitized state has greater capabilities for defense, but it also has a larger field of attack than others. The more reliance on AI increases, the greater its benefits, and with it, the cost of failure rises. Developing countries are less exposed to danger, but they are less capable of resilience on the other hand.

In summary: the individuals, companies, and countries least susceptible to the risk of AI going out of control are those that possess flexibility, independence, and alternatives.

The individuals who are relatively safer are those who use AI without losing their ability to think, work, and make decisions without it. The companies that are safest are those that can revert to human or traditional operation when the system fails and do not grant AI unnecessary permissions.

The safest state is not necessarily the one with the most artificial intelligence, but rather the one most capable of working with AI and working without it when necessary. This means always retaining a human workforce capable of managing core systems if intelligent systems fail.

The most dangerous future scenarios are those in which AI takes over writing, calculating, analyzing, programming, translating, planning, diagnosing, teaching, consulting, and decision-making on our behalf. This means losing control and necessitates a governance system for community immunity to AI; that is, ensuring that individuals, institutions, and states can maximize their benefits from AI without being beholden to it.

There must be empowerment for use, review, rejection, stopping, and continuing without it. Finally, every bank and institution should answer the question: What would happen if the AI systems we rely on disappeared for a week or a month? The answer to this question is sufficient to assess and evaluate the strategic situation of the institution.