The now-famous game of Go between Sedol Lee and AlphaGo has sparked a rapid growth of AI software. However, the development of physical hardware that is able to keep up with the new AlphaGo models was stagnant. In fact, it costs approximately 50,000 USD to run the AI during a single game of Go. As such, in order to commercialize the use of AI, highly efficient hardware development is necessary to keep up with massive data processing power required.
A team of researchers led by Professor Jang-Sik Lee from the Department of Materials Science and Engineering at POSTECH has implemented ferroelectronics to create synapses that resemble analog memory. In recent months, the focus of AI development has shifted to neuromorphic computing, which emulates how neurons process data with our brains. Professor Lee announced that his team was successful in perceiving 90% of number and image patterns with the new invention. He expects the findings to be applicable in many aspects of the industry that make a use of neuromorphic AI techniques.