Scientists at the University of Michigan have developed an artificial intelligence system that could revolutionize the scientific field. The system, called Bacterai, allows robots to conduct up to 10,000 scientific experiments without human interaction, making it a valuable tool in fields such as medicine, agriculture, and ecology.
In a recent study published in the journal Nature Microbiology, the team successfully used Bacterai to study the metabolic processes of microorganisms related to oral health. Led by Professor Paul Jensen, the researchers aimed to determine which amino acids were necessary for the growth of beneficial mouth bacteria, a task made difficult due to the numerous possible combinations of 20 available amino acids.
To tackle this challenge, the Bacterai system was developed. It tested hundreds of amino acid combinations daily and adjusted accordingly based on previous results. Over nine days, Bacterai learned to accurately predict the necessary amino acid combinations for bacteria growth in 90% of cases.
The system’s unique ability to create its own dataset is a breakthrough in AI. Unlike traditional models that rely on pre-existing marked data, Bacterai uses the results of previous tests to determine which experiments will provide the most information. For less than 4,000 Bacterai experiments, most of the rules for proper bacteria nourishment were discovered.
This breakthrough can be applied not only in microbiology, but also in other areas where problems can be formulated as puzzles for AI to solve through trial and error. Bacterai is a significant step forward in using AI to accelerate scientific discoveries. As the study showed, the technology can provide critical insights into the microbial world, including bacteria that affect human health.