Brain-Inspired Memory Device Aims to Boost Artificial Intelligence
Brain-inspired computing has the potential to revolutionize artificial intelligence (AI), surpassing the capabilities of conventional computers, according to a scientist from the University of Hong Kong. Assistant Professor Li Can, from the Department of Electrical and Electronic Engineering, was awarded the prestigious Croucher Tak Wah Mak Innovation Award along with a HK$5 million fund for his research in emerging memory devices, such as the memristor.
Li expressed concerns over the significant time and energy requirements for training AI models using existing computer systems. He noted that if conventional computers continue to be used for AI training, it would necessitate the use of a nuclear power plant to meet the energy demands of the computing system within the next decade. Citing the example of ChatGPT, Li highlighted that it took researchers 36 years with eight processing units to train a GPT-3 model.
Moreover, Li emphasized that conventional computers restrict the development of creative AI as they can only learn from limited databases and follow precise instructions. He argued that they are not an optimal solution for creating advanced AI. To address these limitations, Li’s team has turned to memristors, which mimic the behavior of biological synapses and neurons in human brains, to unlock the potential of training a more power-efficient and human-like AI.
By computing directly within memory, brain-inspired memristive hardware can significantly reduce the time and energy consumed during data transportation compared to conventional computers. Li stated that the computing speed of this hardware could be 100 to 1,000 times faster than the current AI models. However, his ultimate goal is to create an integrated memristor chip that consumes even less energy, akin to a newborn with unique DNA.
Li envisions a future where AI development is internet-free, with AI models running on terminals such as smartphones or even implanted into human bodies to detect diseases. He believes that each chip will have distinct functions, shaping its capabilities and potential applications. To accelerate the computing paradigms, Li plans to recruit more young talents to his team in the next three to five years.
In conclusion, the team led by Professor Li Can is leveraging brain-inspired computing and memristor technology to enhance AI capabilities. Their research aims to achieve a more power-efficient and creative AI system that can operate at significantly faster speeds compared to conventional computers. With the ultimate goal of creating individualized chips, resembling a new life that can think and even sense, Li’s work could pave the way for groundbreaking advancements in AI applications.
(Note: The article contains 446 words)
