Neuromorphic computer systems challenge AI technology
AI and humans in focus
Have you heard of the word neuro-computing? A new field of research is emerging in computer science that could challenge current AI technology. Researchers are looking at how the brain processes information and will create computer systems that will be significantly more energy efficient.
In the human brain, neurons send out impulses to make sense of incoming information, and neuromorphic computing aims to emulate this. The result are novel processors and algorithms that are very energy efficient compared to current AI systems, such as ChatGPT. Neuromorphic computing systems have already been picked up by companies such as Intel or IBM, which have started to create their chips and algorithms.
”We see exponential growth in this technology in the future, even though the research field is new at the moment,” says Jörg Conradt, associate professor of neurocomputing at KTH Royal Institute of Technology.
Based on other computing principles
The challenge with existing computers and intelligent systems is that they use much energy for each – possibly redundant – piece of information. Neuromorphic systems are based on different computing principles, which only use energy when information is updated.
”The more we understand about how the human brain processes information, the more efficient we can emulate it in technology. This includes creating our own processors and rethinking computing primitives. The significant advantage is that we will have the same computing in the future, but it will use much less energy,” says Conradt.
As an example, a research project is underway with Digital Futures and the City of Stockholm to identify who is participating in traffic in Stockholm and how to optimize traveling paths in real-time.
”Current technology to identify whether a car, motorbike, scooter, bicycle or pedestrian is being detected requires a camera and a computer that uses a lot of energy. Our goal is a system that can identify the flows but uses significantly less energy. For traffic planners, this is extremely important information and determines decisions such as where to build an extra lane for cyclists or how traffic signals can be reorganised and adjusted.”
Text: Emelie Smedslund ( emeliesm@kth.se )