Researchers have built a prototype of a computer called a memcomputer that works by mimicking the abilities of the human brain, which simultaneously stores and processes information. The findings pave the way for developing computing machines capable of solving highly complex problems that use interacting memory cells called “memprocessors” to store and process information on a single physical platform.
Although the particular machine presented here suffers from technological limitations, it represents the first proof-of-concept of a machine with this architecture. Current computers store and process data separately, which makes their computing processes very energy-intensive and renders them unable to solve highly complex computational problems that involve large amounts of data. Quantum computing is being explored as a method for solving such problems, but Massimiliano Di Ventra and Fabio Traversa of the University of California, San Diego (UCSD), argue that memcomputers could be a more feasible alternative.
These new, brain-inspired computing devices could even help neuroscientists better understand how the brain operates, they say. In a 2013 paper, Di Ventra and Yuriy Pershin outlined the criteria needed to build a memcomputer. In a follow-up paper published in 2015, Di Ventra and Traversa showed theoretically that memcomputing machines could potentially solve notoriously difficult computational problems, called NP-complete problems, in an efficient way.
In this new study, the authors build a prototype of a memcomputer that solves one such difficult problem. Instead of being shuffled back and forth, data in the memcomputer are directly sent to a computational memory processor (made up of interacting, brain-like memory cells) that is able to compress and preserve information in collective states. The authors also note that the machine they have fabricated in this study is not a general purpose one, but that a more user-friendly version could easily be built with available technology.
Although the particular machine presented here suffers from technological limitations, it represents the first proof-of-concept of a machine with this architecture. Current computers store and process data separately, which makes their computing processes very energy-intensive and renders them unable to solve highly complex computational problems that involve large amounts of data. Quantum computing is being explored as a method for solving such problems, but Massimiliano Di Ventra and Fabio Traversa of the University of California, San Diego (UCSD), argue that memcomputers could be a more feasible alternative.
These new, brain-inspired computing devices could even help neuroscientists better understand how the brain operates, they say. In a 2013 paper, Di Ventra and Yuriy Pershin outlined the criteria needed to build a memcomputer. In a follow-up paper published in 2015, Di Ventra and Traversa showed theoretically that memcomputing machines could potentially solve notoriously difficult computational problems, called NP-complete problems, in an efficient way.
In this new study, the authors build a prototype of a memcomputer that solves one such difficult problem. Instead of being shuffled back and forth, data in the memcomputer are directly sent to a computational memory processor (made up of interacting, brain-like memory cells) that is able to compress and preserve information in collective states. The authors also note that the machine they have fabricated in this study is not a general purpose one, but that a more user-friendly version could easily be built with available technology.