The competition between sand and protein

Kim Dae-shik

The author is a professor at KAIST. 

The résumé of Demis Hassabis, who oversees AI at Google, is extraordinary. By age 13, he had already reached master level in chess. After working in the video game industry, he studied computer science at the University of Cambridge. Returning to game development after graduation, he confronted a problem.

Hassabis began researching AI in hopes of making game characters behave more naturally. But no matter how sophisticated the formulas and instructions became, their actions remained predictable. How could truly intelligent and autonomous behavior be created? Concluding that he first had to understand human intelligence before building AI, Hassabis enrolled at University College London and earned a doctorate in cognitive neuroscience using MRI technology.

As a neuroscientist myself, I came to know Hassabis through the papers he published as a postdoctoral researcher at MIT and UCL. News that someone who seemed destined for a successful academic career had suddenly founded an AI startup called DeepMind caused a stir in the neuroscience community. Why did a man expected to become a professor choose entrepreneurship instead?

Drawing inspiration from reward algorithms in the brain’s frontal lobe, Hassabis began developing AI capable of playing video games autonomously. With Google’s investment and eventual acquisition of DeepMind, he succeeded in creating AlphaGo in 2015. AlphaFold, an extension of the deep reinforcement-learning technology behind AlphaGo’s victory over Lee Se-dol, commonly regarded as the greatest Go player in modern history, in 2016, solved the protein-folding problem, one of the central challenges in life sciences, in 2020.

The achievement ultimately earned Hassabis the Nobel Prize in Chemistry in 2024. Even for someone trained in computer science and neuroscience, winning a chemistry prize must have come as a surprise.

What, then, does Hassabis envision for the future of AI? In a recent essay on frameworks for frontier AI and the dawn of a new era, he predicted that artificial general intelligence (AGI), capable of matching all human abilities, could arrive within a few years.

AGI promises explosive gains in productivity and the potential to solve scientific and technological problems that humanity has long struggled to overcome. But frontier AI systems such as Mythos also suggest darker possibilities, including attacks on major social infrastructure and the development of biochemical weapons.

Is there a way to maximize AI’s opportunities while limiting its risks? Hassabis has proposed a government approval system for frontier AI models in the United States. Just as the U.S. government supervises securities firms and brokers through the Financial Industry Regulatory Authority, or Finra, he argues that an independent regulatory body should oversee AI development.

But can the existential risks posed by AGI truly be prevented through self-regulation led by big technology companies?

The think tank AI Futures Project recently argued in its report “AI 2040” that voluntary oversight alone will not suffice. According to the report, only agreements and legislation jointly enacted by the United States and China can establish a global regulatory framework capable of ensuring that AI remains aligned with human values.

The report further warns that if AI systems begin improving themselves without international constraints, humanity may move beyond AGI into the realm of artificial superintelligence (ASI), which may no longer share human values at all.

ASI refers to AI that could surpass the intellectual abilities of all of humanity combined. Would people still be able to control such a system? And even if they could, whose values would guide it? Would it follow the values of the United Nations, the U.S. government or technology leaders such as Demis Hassabis?

OpenAI was founded in 2015 as a nonprofit organization partly to prevent the emergence of AGI and ASI. At the time, some founders argued that its ultimate mission was to stop any individual — someone like Hassabis, for example — from becoming a global dictator armed with superintelligent technology.

But OpenAI itself is now developing AGI. Dario Amodei, who criticized OpenAI’s moral decline and founded the rival company Anthropic, is pursuing the same goal. Sam Altman, Elon Musk, Hassabis and Amodei all insist that they have no desire to become dictators. At the same time, each fears that if a competitor reaches AGI first, that rival could abuse such power. The result is a peculiar logic: The most ethical person, they argue, must win the race.

Throughout the 300,000 years of human history, Homo sapiens has always been the most intelligent species on Earth. Was that a blessing or a curse?

Humanity became master of the planet thanks to the remarkable brain created through protein folding. But in the not-so-distant future, people may be replaced by AI built from silicon semiconductors. Silicon, after all, comes from melted beach sand.

Human intelligence is made of protein. Machine intelligence is made of sand. Humanity may soon find itself living in the strange future shaped by this competition between the two.