How close are we to AGI?

The development of Artificial General Intelligence (AGI) is believed to be a significant distance away, with estimates ranging from 30 to 50 years or possibly by the end of the century. According to Will MacAskill on the Tim Ferriss Show, AGI is expected to be a transformative technology with substantial potential benefits and risks, comparable to the development of nuclear technology. He emphasizes the importance of being prepared with policies to manage these risks effectively 1.

David Deutsch, discussing AGI on the same podcast, argues that current AI technologies are fundamentally different from what AGI would be. AGI would require unpredictability and the generation of novel ideas, which are characteristics not found in contemporary AI systems focused on specific tasks 2. He also suggests that the development of AGI will require not just technological advances but a new philosophical theory of how such a program should operate, which is still not fully understood 3.

Hence, while AGI remains a concept within reach in the future, it requires significant theoretical and practical advancements beyond current capabilities.