Nobel laureate expects human-level artificial intelligence to emerge in two to three years. What will smart technology be capable of?

At just four years old, Demis Hassabis became fascinated with chess. He was so small that he had to sit on a thick phone book to see the board. By the age of 12, the boy had become one of the strongest chess players in the world in his age category, and he bought his first computer with prize money from tournaments.

Soon, computers interested him even more than chess. At 16, Hassabis was already working at a video game development company and experimenting with early artificial intelligence technologies. His successes were so impressive that his boss offered the young man a million pounds to forgo university and stay with the company. However, Hassabis chose to study at Cambridge, where he pursued computer science.

Later, he became fascinated with studying the human brain and earned a doctorate in neuroscience. Hassabis was interested in how human thinking works and whether a machine with similar abilities could be created. In 2010, he founded the research company DeepMind, where these ideas began to come to life.

Initially, researchers taught artificial intelligence to play old computer games. The systems gradually learned from their own experience and eventually surpassed human players. In 2014, Google acquired DeepMind for $650 million, and two years later, its development AlphaGo defeated one of the world's strongest Go players, an ancient board game considered significantly more complex than chess.

However, Hassabis initially viewed games only as a way to test the capabilities of the new technology. He was much more interested in whether artificial intelligence could solve complex scientific problems. One such task was predicting the structure of proteins, on which biologists had worked for decades.

In 2020, the AlphaFold system created at DeepMind made a breakthrough in this field. It learned to predict the three-dimensional structure of proteins with high accuracy; their shape largely determines their functions in the body. This opened new possibilities for studying diseases and developing medicines.

Later, using AlphaFold, researchers predicted the structures of over 200 million proteins, practically all known to science at that time. These data were made freely available to scientists worldwide. For his work on AlphaFold in 2024, Demis Hassabis, along with his colleague John Jumper, received the Nobel Prize in Chemistry.

Павялічыць

However, the scientist is convinced that the most significant discoveries are yet to come. He hopes that in the future, artificial intelligence will be able to not just assist in solving individual scientific problems, but to make fundamental discoveries independently.

In an interview with The Wall Street Journal, Hassabis stated that so-called Artificial General Intelligence (AGI) — a system with a broad spectrum of intellectual abilities comparable to human ones — could emerge in just two to three years. Earlier this year, he had estimated this timeframe at three to five years.

Will AI Help Conquer Cancer?

For many years, technology companies have promised that artificial intelligence will help humanity conquer cancer. Demis Hassabis believes that now is the time to move from promises to real results. In his opinion, success in treating serious diseases can convince people of the benefits of new technology.

To this end, the scientist founded Isomorphic Labs, a company that uses AI for drug discovery. Its goal is much broader than finding a remedy for a single disease. Hassabis wants to create a system that will fundamentally change the process of discovering new drugs and allow the identification of potential medicines for a multitude of diseases.

Today, the path from a promising substance to a finished drug can take many years. Even if artificial intelligence helps find such substances more quickly, they will still need to be tested for effectiveness and safety. Hassabis understands the complexity of this task but sees no insurmountable obstacles to its resolution.

He hopes that with the development of AI, scientific discoveries will happen much faster, and the development of new treatment methods will become more efficient. However, for now, it's about ambitious plans, not about AI-discovered cures for cancer.

Will AI Be Able to Replicate Einstein's Discoveries?

To assess whether artificial intelligence has truly reached human level and is capable of making scientific discoveries independently, Hassabis proposes an unusual criterion — the so-called Einstein test.

The idea is to train the system only on knowledge humanity possessed before 1901, and then test whether it can independently arrive at Albert Einstein's discoveries of 1905, including the special theory of relativity. If the machine succeeds, it would indicate that it can not just reproduce known information but create fundamentally new scientific ideas.

Hassabis hopes that in the future, such systems will help researchers discover fundamental laws of physics and better understand the nature of reality.

Павялічыць

How to Make Powerful AI Safe?

Despite his optimism, Hassabis acknowledges that the development of artificial intelligence carries serious risks. In terms of its impact on humanity, he compares this technology more to the discovery of fire or electricity than to the advent of the internet. To mitigate potential danger, the scientist proposes creating a special organization that will vet the most powerful AI systems before their public release. It should be funded by the technology companies themselves but operate under government oversight.

At the same time, Hassabis warns against hastily imposing restrictions. Artificial intelligence is developing so rapidly that regulations could become outdated before they even take effect. Therefore, the scientist proposes a flexible control system that can be adapted as technology evolves. In his opinion, incorrect regulation could do more harm than its absence.

Hassabis believes that the main task for the coming years is not only to create human-level artificial intelligence but also to ensure its safe use for the benefit of all people. He urges discussions now on how the emergence of AGI will change society and what needs to be done to make its advantages accessible to everyone.

The scientist himself hopes that in the future, together with powerful AI systems and leading researchers, he will be able to seek answers to fundamental questions of physics and better understand the nature of the Universe.

• 5
• 2
• 0
• 1
• 0
• 0
• 0
• 0