In the vibrant landscape of artificial intelligence (AI) research, Hong Kong scientists embark on an ambitious mission—seeking inspiration from the human brain to enhance AI capabilities.
Spearheading this revolutionary quest is Li Can, an assistant professor at the University of Hong Kong, leading a team aiming to propel AI systems into realms of lifelong learning and versatile task performance, mirroring the astounding capacities of the human mind.
"At the core of our pursuit lies the aspiration to develop AI systems capable of perpetual learning and adaptive functionality, reminiscent of the human brain's unparalleled capabilities," Li asserted. Central to this endeavor is the quest for more efficient and potent hardware, a vital enabler for the next generation of AI systems.
Li Can elucidates the distinctive strengths of both realms. While human brains excel in experiential learning and reasoning from ambiguous data, computers showcase prowess in scientific computations and repetitive operations.
To bridge this gap, Li and the team have turned their gaze toward an emerging memory device—the memristor. This advanced microelectronic platform mimics the intricate behaviors of biological synapses and neurons, offering a tantalizing glimpse into revolutionizing AI hardware.
While acknowledging the human brain's supremacy in experiential learning and reasoning based on ambiguous data, Li emphasized computers' forte in scientific computations and pattern-driven derivations. Each chip conceived in this quest mirrors a newborn, akin to a training model poised to evolve over time, akin to a growing child.
The brain-inspired devices, veering away from traditional computers, promise a paradigm shift in computational technology, offering promising strides in accelerating graphics processing units (GPUs).
With AI models scaling beyond trillions of parameters, the memristor emerges as a crucial element, enabling direct computation within memory—eliminating data transfer bottlenecks akin to the brain's localized information processing.
Energy-efficient memristor chips transforming AI tasks
Li envisions a future where energy-efficient memristor chips could transform AI tasks, migrating from energy-intensive data centers to everyday devices like smartphones and watches. These chips hold promise in diverse applications, spanning health-monitoring implants and expediting virus genome sequencing, potentially revolutionizing disease detection and genome analysis.
The recent accolade of receiving funding from the Croucher Tak Wah Mak Innovation Awards serves as a testament to Li's pioneering efforts. The financial boost will fortify his team's research endeavors, fostering innovation and attracting top talent to Hong Kong's scientific landscape.
In Li's vision, Hong Kong stands at the forefront of scientific exploration, empowered by talent and pivotal financial support—an ecosystem propelling transformative breakthroughs at the intersection of AI and neuroscience.
Originally published on Interesting Engineering : Original article