When do you think AI will fully surpass human intelligence?

On this question, people hold completely different views, much like discussing "when we will be able to travel interstellarly"—some believe it's just around the corner, while others think it's still a long way off.

First, we need to break down the concept of "intelligence."

In certain "individual tasks," AI has actually surpassed humans long ago. For example:

1. Computation and Memory: There's no comparison in this regard. The computational speed and information storage capacity of any ordinary computer far exceed those of the strongest human brain. We can no longer live without calculators and search engines.
2. Specific Skills: In many domains with clear rules and objectives, AI performs better than most people, and even better than experts. For instance, AlphaGo in Go, today's leading AI painting tools, or AI systems that analyze medical images and detect early-stage lesions. In their respective "specialties," they have already reached superhuman levels.

But you're asking about "overall" surpassing humans, which touches upon the most critical challenge, namely, Artificial General Intelligence (AGI).

Artificial General Intelligence refers to AI that, like humans, possesses common sense, can reason, flexibly adapt to new environments, be creative, and even have self-awareness and emotions. This is completely different from the "specific skills" mentioned above.

A three-year-old child knows that an apple in their hand is edible, but a mobile phone is not; seeing dark clouds, they would infer that it might rain. This seemingly simple common sense and reasoning ability is an enormous challenge for current AI. AI needs to learn astronomical amounts of data to "summarize" some patterns, whereas a human child can generalize from just a few examples.

Therefore, regarding the timeline for "overall surpassing humans," there are currently several main schools of thought:

• Optimists: Some leading tech figures and futurists (like Ray Kurzweil) believe that this "singularity" will arrive soon, possibly around 2045. They believe technology will accelerate rapidly and quickly overcome current limitations.
• Cautious View: Many neuroscientists and AI researchers believe it's not that simple. Because we haven't even figured out how the human brain generates consciousness and intelligence, it's even harder to create one from scratch. They think it might take a century, or even longer.
• Skeptics: Still others believe that machines may never truly possess consciousness and understanding like humans. AI's "intelligence" and human "intelligence" might not be the same thing at all. It might perform better than humans on all tasks, but it still lacks true "understanding" and "feelings."

Overall, I think instead of trying to guess a specific year, it's better to focus on the present. AI technology is integrating into our lives at a visible pace, and it's like a new "species" beginning to coexist with us. How to collaborate with it, how to leverage it to solve real-world problems (such as conquering diseases, addressing climate change), and how to establish rules to prevent its misuse—these are perhaps more important and urgent questions than "when it will fully surpass us."