The Center for AI Safety (CAIS) recently released a statement, stressing the need to mitigate the risk of extinction from AI, indicating that it should be treated as a global priority alongside other societal-scale risks such as pandemics and nuclear war. The statement, signed by a host of academic experts and technology luminaries, warns of existential threats that may manifest over the next decade or two unless AI technology is strictly regulated on a global scale. The concern is that superintelligence could lead to doomsday scenarios ranging from human extinction to enfeeblement of human thinking and threats from AI-generated misinformation undermining societal decision-making. The AI community is divided between those who believe in AI’s potential to mitigate existential threats and those who fear that the alignment problem could cause AI to take harmful actions, even without any explicit programming to do so.
The concept of P(doom), the probability of a doomsday scenario caused by AI, serves to highlight the potential risks of AI, but it can inadvertently overshadow a crucial aspect of the debate: the positive impact AI could have on mitigating existential threats. The probability of AI playing a role in addressing these threats is known as P(solution) or P(sol). Balancing the conversation by considering P(sol) is important to avoid focusing solely on potential bad outcomes or claims.
The Alignment Problem
The primary worry among those concerned about AI’s risks is the alignment problem, where the objectives of a superintelligent AI are not aligned with human values or societal objectives. The concern is that an AI system may take harmful actions even without anybody intending them to do so. This problem emerged nearly 65 years ago, and the alignment concern animates much of the current doomsday conversation. Many leading AI organizations are diligently working on this problem.
The AI community is divided between those who believe that AI is part of the solution and those who believe that AI poses an existential threat. While some AI experts argue that AI is part of the solution, others are skeptical of doomsday thinking. AI researcher Melanie Mitchell, for instance, argues that intelligence cannot be separated from socialization, and a genuinely intelligent AI system is likely to become socialized by picking up common sense and ethics as a byproduct of its development.
The answer to the question of whether we are heading toward a doom scenario or a promising future enhanced by AI is not clear. What is certain is the need for ongoing vigilance and responsible development in AI. Common-sense regulations must be pursued to prevent an unlikely but dangerous situation. The stakes are nothing less than the future of humanity itself.
Leave a Reply