Researchers at the USA’s Defense Advanced Research Projects Agency want to work out how computers can mimic a key portion of our brain.
The group called for information on how it could develop systems that solve “extraordinarily difficult recognition problems in real-time”.
According to Network World, systems presently only offer partial solutions to this problem.
They can’t scale to larger, more complex datasets, and are “compute intensive, exhibit limited parallelism, and require high precision arithmetic”.
DARPA wants to mimic the neocortex which is used in higher brain functions such as sensory perception, motor commands, spatial reasoning, conscious thought and language.
The idea is to develop a “Cortical Processor” based on Hierarchical Temporal Memory.
Its “call for information” said that computing was at a point where some basic algorithmic principles had been spotted and merged into machine learning and neural network techniques.
These algorithms were inspired by neural models, in particular neocortex, and can recognise complex spatial and temporal patterns and can adapt to changing environments.
DARPA added that the cortical computational model should be fault tolerant to gaps in data, massively parallel, extremely power efficient, and highly scalable. It should also have minimal arithmetic precision requirements, and allow ultra-dense, low power implementations.
Overall, the new RFI will be part of the research and development DARPA has been doing to a computer with similar form and function to the mammalian brain. Such artificial brains would be used to build robots whose intelligence matches that of mice and cats, DARPA says.