The project, announced earlier this week, wants to bring artificial intelligence closer to reality by gaining a greater understanding of how human learning operates. As Engadget reports, three departments at Harvard University have received a total of $28 million from the U.S. government’s Intelligence Advanced Research Projects Activity (IARPA) to investigate how the brain learns new things so quickly.
The project aims to record activity in the brain’s visual cortex and map the connections it makes while learning. The researchers will be analysing the brain in “unprecedented detail” with the aim of generating a petabyte of data – 1000 terabytes, or enough to fill 1.6 million CDs – that can then be used to create computer algorithms capable of recreating the brain’s learning process.
The investigation was designed in response to the vast amounts of data that intelligence agencies are now faced with every day. Humans can’t make sense of the huge volumes of data involved, despite possessing much greater pattern-recognition abilities than even the most powerful machine learning systems built to date.
IARPA set Harvard University the challenge: “figure out why brains are so good at learning, and use that information to design computer systems that can interpret, analyze and learn information as successfully as humans.”
The project will begin by training rats to recognise visual objects displayed on a computer screen. The team of researchers will record the activity of the rats’ visual neurons using next-generation laser microscopes developed specially for the project.
A “substantial portion” of the rat’s brain, one-cubic millimetre in size, will then be dissected into ultra-thin slices and imaged under the world’s first multi-beam scanning electron microscope. This will detect changes in the brain’s structure as a result of learning new information, helping the researchers to understand how the connections form.
David Cox, leader of the project and assistant professor of molecular and cellular biology and computer science at Harvard, describes the experiment as a “moonshot challenge.” He said: “The scientific value of recording the activity of so many neurons and mapping their connections alone is enormous, but that is only the first half of the project. As we figure out the fundamental principles governing how the brain learns, it’s not hard to imagine that we’ll eventually be able to design computer systems that can match, or even outperform, humans.”
Once the first stage has been completed, the project will continue to investigate how the brain uses its connections to quickly process information and detect patterns in the things it observes in the world. The results of the study will directly influence future developments in computer science and machine learning, creating artificial intelligences with more similarities to human life than ever before.
The team thinks the technology could prove vital to developments in areas including self-driving cars, future antivirus software capable of protecting whole networks and the automatic analysis of MRI scans. The possibilities could be almost endless though as computers become able to replicate the brain — still the most sophisticated pattern-detector around — more closely than ever before.
