The field of artificial intelligence is receiving more attention than ever before. As discussions about AGI (Artificial General Intelligence) become increasingly active, both excitement and concerns about the changes this technology will bring to humanity are growing simultaneously. Through a recent interview with Demis Hassabis, CEO of Google DeepMind and Nobel Prize winner, we gained deep insights into the present and future of AGI.
AGI Is Closer Than We Think
Hassabis revealed that when he founded DeepMind, he set a 20-year mission to "solve intelligence, and then use that intelligence to solve everything else." Remarkably, he assessed that the current plan is progressing almost exactly as expected. Even more noteworthy is his belief that there's a 50% probability of achieving AGI within the next 5-10 years.
However, the definition of AGI is crucial here. Hassabis defines AGI as "a system that can demonstrate all cognitive abilities that humans possess to some degree." It's not simply about showing excellent performance in specific fields, but about being able to generalize across all domains. The current inconsistency of AI systems—solving International Mathematical Olympiad problems at gold medal level while making mistakes on high school math problems or counting letters in words—perfectly illustrates these limitations.
Geopolitical Competition and the Safety Dilemma
As AI development competition intensifies, "who gets AGI first" has emerged as a national-level issue. Some propose a "hard takeoff" scenario where the country that develops AGI first will gain absolute dominance. However, Hassabis took a cautious stance on this.
He acknowledged that while AGI systems might improve themselves and achieve rapid development, a gradual development scenario is also entirely possible. More importantly, he emphasized that such systems inevitably reflect the values of their designers and the cultures they belong to. Therefore, who develops AGI, where, and with what values will have profound implications for all of humanity.
Interestingly, unlike a few years ago when AI companies were calling for regulation, the current U.S. government is more interested in accelerating AI development to win the competition with China. While Hassabis still advocates for reasonable and smart regulation, he emphasized that such regulation must be international. Due to the nature of digital systems, regulating only one region doesn't guarantee global safety.
Job Changes: Destruction or Evolution?
If AGI can do everything humans can do, what happens to jobs? Hassabis assessed that AI has been playing a complementary role so far. Tools like AlphaFold are helping accelerate scientists' research, but they're working to increase productivity rather than replace jobs.
He optimistically predicted that better jobs utilizing new technology will emerge, just as when the internet or mobile technology appeared. Using healthcare as an example, he predicted that while AGI might help with technical tasks like diagnosis, human aspects like nursing care will still be handled by people.
For current graduates, he advised deep immersion in these new tools. He recommended studying STEM or programming to understand how these systems are built and developing the ability to modify open-source models in their own way. He emphasized that the next 5-10 years will be the most productive period, and those who naturally handle these tools could see their productivity increase tenfold.
The Era of Radical Abundance Is Coming
Hassabis's vision of the post-AGI future is quite optimistic. Using the term "era of radical abundance," he predicted that AGI will solve the core problems humanity faces. Examples include healthy longevity through disease treatment, development of new energy sources, and better batteries and superconductors.
The example of solving water scarcity was particularly impressive. Currently, desalination can only be utilized by wealthy countries due to energy costs, but if clean, nearly free energy like fusion were available, water scarcity would be solved. He believed that when such fundamental abundance is realized, humanity can escape zero-sum games and enter a true era of prosperity.
However, he acknowledged that humanity is currently not skilled at cooperation. Even with climate change, we know the solutions but aren't implementing them. Fortunately, he expected the situation to be different because the radical abundance that AGI brings won't require people to make sacrifices.
Realistic Concerns and Responses
Of course, it's not all rosy. Hassabis pointed out two major risks. First is individuals or rogue states misusing AI technology for harmful purposes, and second is the technical risks of AI itself. As AI becomes more powerful and autonomous, whether we can properly establish safety measures becomes crucial.
To address these concerns, DeepMind is investing more resources in AI safety and cybersecurity. They're also focusing on research into AI system controllability and interpretability. Hassabis believed that technical problems can be solved through sufficient time, caution, and scientific methodology, but geopolitical problems might actually be more challenging.
Backlash Against Change and Understanding
He also addressed people's anger and backlash against AI development. Hassabis said this reaction is similar to the Industrial Revolution and completely understandable. Since this change will be at least as significant as the Industrial Revolution, and probably much larger, it's natural to be scared.
However, he emphasized that through concrete achievements like AlphaFold, they're showing that AI can actually help medicine and drug development. If we can treat terrible diseases that torment families, not doing so would actually be immoral.
Assessment from an Expert's Perspective
Synthesizing Hassabis's interview, he presents a balanced view of technological optimism and realistic concerns. While the faster-than-expected progress in AGI development is noteworthy, he also emphasizes the importance of safety and international cooperation.
What's particularly impressive is that he's not simply absorbed in technology development but considers social, economic, and geopolitical contexts together. There's a recognition that the changes AGI will bring could determine the future of all humanity, beyond just technical issues.
However, there are some points to consider critically. First, seeing a 50% possibility of achieving AGI within 5-10 years is quite an aggressive prediction. Given the current limitations of AI systems, it's questionable whether such radical development is really possible. Second, even if an era of radical abundance comes, whether it will be distributed fairly is a separate issue. Even now, there's technically sufficient abundance, but many people suffer due to distribution problems.
What We Need to Prepare
Synthesizing Hassabis's advice, at the individual level, it's important to develop the ability to actively adapt to and utilize new AI tools. Beyond simply learning how to use them, we need the ability to understand how these systems work and creatively apply them to our own fields.
At the societal level, establishing AI governance and international cooperation systems is urgent. As the pace of technological development accelerates, safety measures and regulatory systems must develop together. International coordination is essential, especially given that AI is a technology that affects the entire world.
Economically, we need new economic theories and institutions suited to the AGI era. As Hassabis pointed out, it's time for economists to participate more actively in this issue. Reforming education systems and building social safety nets to prepare for job changes are also important tasks.
In summary, we may be standing at the threshold of the greatest change in human history. If the future Hassabis envisions becomes reality, it will certainly be remarkable. However, the risks and challenges that may arise in the process are also significant. Along with expectations for technological progress, careful preparation and international cooperation are more important than ever.