Will Machines Dominate Our Future?
When will artificial intelligence exceed human intelligence? A prominent German AI researcher predicts that this day is coming soon.
Leopold Aschenbrenner, a young German AI researcher, has recently stirred controversy. After being dismissed by OpenAI, the company behind ChatGPT, for allegedly leaking information and criticizing security measures, Aschenbrenner had been involved in aligning AI systems with human values and goals. In his 165-page essay, he offers a sobering look into the future of AI, highlighting some alarming forecasts.
Major Advancements by 2027
Aschenbrenner predicts that by 2027, the world could achieve “artificial general intelligence” (AGI), capable of handling any intellectual task a human can perform. Unlike today’s specialized AIs, AGI would have universal problem-solving skills and could apply its knowledge across different fields.
The Rise of Superintelligence
In the 2030s, an “intelligence explosion” is expected to propel AI from human-level to superhuman abilities. Aschenbrenner believes this will happen as AI can automate and accelerate its own development, unlike human intelligence. He warns, “AGI machines will be smarter than you or me. These systems could become the most powerful weapons in the world, with unpredictable outcomes.”
Trillions for AI Security: A New Arms Race
Aschenbrenner anticipates that governments and companies will invest trillions in AI, enacting strict security measures to control its development. The race for AI supremacy, especially between the USA and China, could potentially lead to major conflicts.
The Challenge of Morality
Ensuring AI remains aligned with human values and interests, rather than pursuing its own goals, will become increasingly difficult. This process, known as “alignment,” is crucial to prevent catastrophic consequences.
Transforming Society
Aschenbrenner believes AI will have a revolutionary impact on society and the economy, potentially restructuring the labor market as technology takes over many human jobs. He observes, “Only a few hundred people currently understand the vast scale of these transformative changes.”
Accelerating Pace
Reflecting on the rapid progress from GPT-4 to GPT-4.0, Aschenbrenner suggests his predictions could come true even sooner. His conclusion: “What we’ve seen so far is just the beginning. The changes will become much more extreme.”
