The discovery of new materials has always been the driving force for the advancement of various industries. However, the development of new materials is now facing three major bottlenecks: it requires a lot of time, cost, and labor.
The development of new materials has always relied on researchers to continuously experiment and find the best solution through different experimental combinations. Just like the story of how Thomas Edison brought electric light from the laboratory to every household — after thousands of experiments and failures, he finally found a material that could make a light bulb last for a long time.
This trial-and-error method may have worked in the past, but what if the experimental variables grow from three to thousands? As material components become more complex and synthesis techniques more refined, the number of experiments required increases exponentially, making the development of new materials incapable of keeping up with the pace of industrial progress.
“Eroom’s Law” of new drug development is the best example to explain this dilemma. In the semiconductor industry, Moore’s Law states that the number of transistors per integrated circuit doubles approximately every two years; Eroom’s Law, on the other hand, refers to the increasing complexity of developing new drugs. Instead of decreasing over time, the cost of new drug development doubles approximately every nine years.
According to the latest research, the cost of developing a new drug until it reaches the market has reached approximately US$1.3 billion. Similarly, the cost of optimizing and developing new materials is as high as tens of millions or even hundreds of millions of dollars, and the time required is at least “years.”
But now, with a combination of AI and robots, we see more possibilities in the development of new materials.
At the end of 2023, Google’s AI company, DeepMind, published a paper in the well-known journal “Nature” stating that it had discovered 2.2 million new crystals through the AI tool “GNoME,” which would have required nearly 800 years of human research in the past. Earlier this year, Microsoft announced that a scientific team used Microsoft’s AI tools to discover potential new materials for lithium batteries, from more than 30 million different materials, in just 80 hours. Without AI, this research may have taken 20 years.
Another trend that has recently emerged in the field of materials research and development is “Self-Driving Labs” (SDLabs).
SDLabs not only uses AI to simulate material synthesis but also integrates robots to automatically perform experiments following AI’s strategies. The reason for robot integration is that the properties of certain materials still need to be observed and verified through actual experiments, and thus it is difficult to completely replicate these properties through computer simulations alone. This not only allows SDLabs to significantly shorten the required time for material development; it also frees researchers from repetitive tasks, allowing them to focus on higher-level work.
At present, SDLabs has been actively introduced into industries such as pharmaceuticals, biotechnology, and materials science. For example, Japan’s largest pharmaceutical company, Takeda Pharmaceuticals, and IBM Research are both cooperating with Atinary, a new company that develops SDLabs. A professor from the Laboratory for Soft Materials at MIT even described Atinary’s SDLabs as the “iPhone of R&D.” It is easy to use, and it helps researchers complete experiments in just one week that would have otherwise taken two years to conduct.
AI allows humans to conduct research and achieve innovation in a more efficient, more cost-effective, and more systematic way. However, looking back at the history of human inventions, some of the most representative inventions stem not only from systematic research but also from human curiosity, keen observation, and serendipity.
For example, antibiotics were discovered only because the researcher chose to look further into a contaminated petri dish that was supposed to have been thrown away, and accidentally discovered antibiotics that could kill bacteria.
Such serendipity and uncertainties are the most challenging and yet the most fascinating aspect of all innovations. AI may not be able to replicate the curiosity or the observational ability of a human being. However, just as with all scientific research, as long as we continue to keep an open and curious mind when it comes to AI, more unexpected gains are sure to be unlocked, one by one!
