An interdisciplinary, interuniversity group of scientists from Vanderbilt, Cornell, and CFD Research Corporation have created an artificial intelligence capable of solving complex scientific problems from scratch. The AI, called ABE (Automated Biology Explorer), “discovered†how glycolysis produces energy in a living cell by looking at a set of data and then squeezing it into a mathematical formula, just like a human biology researcher.
ABE is powered by the freeware Eureqa software, and before tackling a task the only functions it can perform are addition, subtraction, multiplication, and division. Using these fundamental calculations, ABE can look at the results from an experiment and derive the mathematical formulae that explain biological and chemical processes a priori, usually in a fraction of the time that it would take a fleshy, brain-encumbered scientist to do arrive at the same conclusion.
Therein lies the crux: computer chips and robots are perfectly suited to scientific research. They can’t (yet) make an instinctively, inspired leap towards “eureka!â€, but they can parse terabytes of information and perform billions of calculations in the same time that you or I could work out a single differential equation.
“Biology is more complex than astronomy or physics or chemistry,†says John Wikswo, a Vanderbilt professor who worked on ABE. “In fact, it may be too complex for the human brain to comprehend.†The truth is, in the sphere of biological sciences, the gap between theory and data is vast. We are constantly accruing more and more data from experiments, but because of the massive variety and complexity of even a single animal cell, we just can’t keep up. Thus the need for automation in the form of ABE.
The next step is to give ABE a physical, real-world presence in the form of a microformulator, alaboratory-on-a-chip (LoC) that uses microfluidics to test cell metabolism. Much like the silicon chip that can sequence the human genome in just a few hours, ABE — perhaps with a robotic actuator or two — will then be able to analyze and process data much faster than its creators, and hopefully some novel discoveries will quickly follow.
Then it’s just a matter of strapping more and more LoCs onto ABE until he’s a veritable whirlwind of laboratorial, scientific endeavor that churns out new findings every few seconds… and then, unless scientists belong to a union of some kind, it will be the robotic pork ham deboning debacle all over again.
Source : http://www.extremetech.com/extreme/100195-ai-solves-complex-biology-problem-from-scratch
