DARPA’s Latest Idea Could Put Today’s Turing-era Computers At Risk
DARPA is known for creating breakthrough technologies, now indulged in Artificial Intelligence projects. Its latest idea is to create intelligent computers (digital beings) that are always learning and adapting, much like humans.
Thus, computers, should be made like humans, DARPA believes. The defence agency wants to develop computers that can think and act like biological beings and improve itself with experience. Therefore, according to DARPA “That’s a far cry from what even simple biological systems can do as they adapt to and learn from experience.”
Hence, DARPA bringing its new Lifelong Learning Machines (L2M) program to develop next-generation ML(Machine Learning) technologies which learn new situations. It adapts it to become better and more reliable, doing tasks within predefined limits that the system cannot override.
DARPA’s Latest Idea
“Life is by definition unpredictable. Hence, it is impossible for programmers to anticipate every problematic or surprising situation that might arise, which means existing ML systems remain susceptible to failures as they encounter the irregularities and unpredictability of real-world circumstances,” said L2M program manager Hava Siegelmann. He also said regarding ML system’s ability. Its ability to perform in a new kind of situation, by taking the system out of service. Then add it with more data sets relevant to that new situation, is not scalable.
Thus, an adaptive computer that draws on experience to make decisions has been a “long-standing” goal, said Hava Siegelmann, program manager for the L2M project at DARPA.
However, today’s computers are designed to run on prewritten programs incapable of adapting as they execute. Because a model hasn’t changed since the Alan Turing, British polymath developed the earliest computing machines in the 1930s,” said Siegelmann, who has a background in neuroscience.
L2M program will last for four year aims to adapt “lessons learned” to new situations and to control and monitor ML behaviour.
DARPA will host a Proposers Day on March 30, 2017, at the DARPA Conference Center.