If you asked a doctor, it’s very likely that they would say that there is no way. But that’s because understanding why all diseases might be solvable has to do with physics, computers, and the nature of intelligence and less with the practice of medicine itself.
I’m using the word solvable and not curable because there is a difference between a solution to a problem and actually hands-on solving it. To illustrate, the procedure for a successful surgery isn’t the same as actually doing a surgery.
The person claiming to be able to solve all diseases by 2035 is Demis Hassabis. To understand why he does so and whether it’s a plausible claim, we have to get to know Demis and his goals first.
Demis’ life goal has been to understand the nature of reality; he’s on a search for truth, to find answers to the big questions. What’s time and consciousness? What is the meaning of it all? And how is it that our brains can understand and figure out things in the first place? He has thought about questions like that for 30+ years, and they have been his main driver on his path through life.
He’s asking the questions physicists have been asking forever and he believes in a multidisciplinary approach, experiencing several disciplines in depth to find useful concepts that he can use to answer his questions.
From being the captain of the England chess youth team, to designing and programming video games, becoming a computer scientist, earning a PhD in cognitive neuroscience, founding DeepMind, and finally winning the Nobel Prize in chemistry last year, to say the least, he seems to creep closer to his goal.
Today, most people think of Demis as a CEO and artificial intelligence researcher. One with really strong beliefs in what AI will be able to achieve. He believes that AI will allow him to reach his goal: building a computer system so intelligent that it can answer his questions for him.
His PhD in neuroscience gave him the existence proof of intelligence in the human brain. He then founded DeepMind in 2010, with an ambitious mission, which at the time was stated as Step 1: solve intelligence; Step 2: use it to solve everything else. DeepMind gave themselves a 20 year timeline to create such a system that would be capable of solving everything else. Today, we are 15 years into this journey.
DeepMind has been able to keep up with its schedule. In the past years, they have created AlphaZero, AlphaFold, AlphaEvolve, Gemini, one of the most accurate weather forecasting model in the world called GraphCast, and a few others. And then, very recently, Demis announced that he believes that by 2035 all diseases could be solved. For that, he created Isomorphic Labs with the mission statement: Solve all disease.
I believe that this is highly probable.
To achieve it, Demis and his companies must stay focused on his long-time goal. This is crucial considering he is probably the only one who has thought about this for 30+ years and who has been constantly achieving milestones. He probably has the best intuition of all people on earth, on what direction to take and what to do next. If someone is on the right path, then it’s probably him. So he must keep pushing forward.
Today’s AI systems can already go beyond what anybody teaches them. Meaning, they express some level of creative behaviour. They have some imagination. But their level is limited. Demis describes 3 levels of creativity:
Level 1: Interpolation: for example, averaging images of cats and drawing a cat.
Level 2: Extrapolation: for example, AlphaGo, move 37.
Level 3: Invention: it could invent the game of Go or invent general relativity with the same information Einstein had at the time.
At the moment, no AI system is at level 3. However, it’s not sure that level 3 of creativity is needed to “solve all disease.” It seems more likely that level 2 is enough, the same level that was needed to solve protein folding.
It is true that a statement such as “Solve all disease” is useful to attract talent. And since Isomorphic Labs just raised 600 million, they are looking for the best talent in the world to go after their mission. However, given Demis’ history and achievements, he seems to be very accurate in his statements and predictions. On top of that, he already has access to all crucial resources: money and talent from DeepMind and Google, and support from Sergey and Larry, the two cofounders of Google who still today have absolute voting power at Google, which DeepMind is a part of. And they are 100% behind Demis.
There is one potential obstacle, however, which is simply bigger than anything Demis and his companies might be able to overcome. Namely, that the nature of the universe itself simply doesn’t allow them to solve all diseases. This has to do with the nature of the AI systems, which types of problems they are able to solve, and with the nature of complexity in the natural world.
For a problem to be solvable by today’s AI systems, the problem has to have a “massive combinatorial search space, a clear objective function (metric) to optimise against, and either lots of data and/or an accurate and efficient simulator.” Demis argues that every problem found in nature is such a problem. He believes that nature is full of patterns and structures that these AI systems will be able to recognize. This would include human health, which includes all diseases. But he of course has no proof for that and bases this belief on what he has been able to observe in his work and through the problems his previous AI systems have been able to solve.
I believe that he is right about being able to “solve all disease” because I believe in his view of what the best explanation is of what the world is made of, namely Information (more about that in my next article). I do not, however, have any idea about the timeline of when it will happen. I simply do not know enough about the actual happenings at AI companies. But if someone in this industry has to be taken seriously when mentioning goals or timelines, then that’s Demis.