Penn Researchers Develop New Chip That Uses Light for Faster, More Efficient AI Computations
• Researchers at Penn have developed a new silicon photonic chip that performs complex math for AI using light instead of electricity, allowing for faster processing speeds and lower energy consumption.
• The chip brings together research on manipulating materials at the nanoscale to control light propagation with existing silicon photonic platforms used to mass produce chips.
• The chip performs vector-matrix multiplication, an essential mathematical operation for neural networks and AI systems.
• The chip could potentially accelerate training and classification in applications like graphics processing units.
• Using light for computation also provides privacy advantages since sensitive information doesn't need to be stored in a hackable memory.