In a recent DeepMind podcast, co-founder Shane Legg shared his views on the development of artificial intelligence. He believes there is a 50% chance of achieving "minimum AGI" within the next two years. Minimum AGI refers to an artificial intelligence system capable of performing most cognitive tasks typically carried out by humans, while "full AGI" encompasses the entire range of human cognition, including remarkable achievements such as developing new scientific theories or composing symphonies.

AI robot at work

Image source note: The image was generated by AI, and the image licensing service is Midjourney

Legg pointed out that after achieving "minimum AGI," "full AGI" is expected to arrive within three to six years. He proposed a comprehensive testing system to evaluate the progress of artificial intelligence. If an artificial intelligence system can pass all typical human cognitive tasks and even after months of rigorous testing, no weaknesses in the system can be found by a human team, then this goal is considered achieved.

This statement has sparked widespread discussion about the future of artificial intelligence, especially in the context of rapidly advancing technology, where the potential of artificial intelligence seems to be getting closer to human cognitive levels. Legg's framework offers a new perspective for understanding the evolution of artificial intelligence, particularly in the transition between "minimum AGI" and "full AGI."

With the rapid development of technology, experts' opinions within the industry are also evolving. Although Legg's predictions are challenging, they also motivate research teams to continue their efforts and drive progress in the field of artificial intelligence. The developments in the coming years will be critical, as everyone is watching how this process will change our lives and work methods.

Key points:  

🔍50%: Legg predicts a 50% chance of achieving "minimum AGI" by 2028.  

📅 Time frame: "Minimum AGI" is expected to be achieved within two years, and "full AGI" is expected to follow within three to six years.  

🧪 Testing criteria: Legg proposed a comprehensive testing system to ensure that AI systems can pass all human cognitive tasks.