“Step one: Solve intelligence. Step two: Use it to solve everything else.”
Demis Hassabis has summarized his organization’s mission in two steps: “Step one: Solve intelligence. Step two: Use it to solve everything else.” The formulation is intentionally audacious. It treats intelligence not only as a subject to understand but as a general instrument for discovery. If researchers can build systems that learn, reason, plan, and transfer knowledge across problems, those systems could help scientists navigate spaces too large for unaided trial and error. Advances in games and pattern recognition then become testing grounds for methods that might contribute to biology, materials, energy, mathematics, or other complex domains.
The success of AlphaGo illustrates the first half of that strategy. Mastering Go required evaluating an enormous number of possible positions and discovering moves that surprised expert players. The result did not mean the system understood every form of intelligence, but it showed that learning and search could produce strategies beyond established human play. Later work on protein structure demonstrated how related methods could accelerate a scientific task with practical significance. These achievements support the idea that better computational intelligence can become a multiplier for research rather than merely a substitute for routine labor.
“Everything else,” however, contains many problems that are not solved by prediction alone. A model may identify a promising molecule without deciding how a treatment should be priced or distributed. It may optimize energy use without resolving conflicts over land, jobs, or public priorities. Scientific breakthroughs enter institutions shaped by law, economics, culture, and unequal power. The more capable the system, the more important it becomes to distinguish technical discovery from legitimate social decision making. Intelligence can expand available options, while people remain responsible for deciding which options are just and whose needs come first.
Hassabis’s quote is best read as a research horizon joined to a governance challenge. AI may help compress the search for solutions, but it does not automatically define the problem correctly, guarantee reliable results, or distribute benefits fairly. Progress requires open scientific evaluation, careful validation in real settings, and safeguards against dangerous use. It also requires public investment so discoveries serve more than the institutions able to purchase the most computing power. Solving aspects of intelligence could unlock extraordinary tools. Using them to solve human problems will still demand human values, cooperation, and accountability. The two steps are therefore connected by a third, continuing task: decide responsibly what counts as a problem worth solving. That work cannot be outsourced to an optimizer, because it involves meaning, fairness, and commitments among people who may experience the result very differently.
Games provide clear rules and feedback, making them useful environments for testing learning, search, and planning.
Breakthroughs there can reveal techniques that later contribute to less structured scientific problems.
Systems can evaluate vast spaces of candidate structures, strategies, or experiments faster than unaided researchers.
Their value grows when outputs are validated through domain expertise, reproducible evidence, and real-world testing.
Technical knowledge cannot determine fair distribution, acceptable risk, or legitimate public priorities by itself.
Human institutions must choose which problems to pursue and remain accountable for how discoveries are used.
Explore more "Quotes of The Day"
Continue exploring how writers, thinkers, artists, and advocates have described the choices that shape an everyday life.