With AI moving to the edge, Arm's Paul Williamson talks about the move to small language models; gaining efficiency without losing performance; and how software devs are transforming hardware.
General purpose AI models are great if you want to discuss everything from geopolitics to baking with the same chatbot; but when you have hard limits on data and compute, you need a more focused system. Enter, small language models.
This week Tom talks to Arm's Paul Williamson about what the rise of SLMs means for edge computing and the people who use it; different ways to shrink language models; and how a change in software dev behaviour is forcing hardware makers to evolve.
Become a member of Computing today at www.computing.co.uk
LinkedIn: @computing-- / Tom Allen
Twitter: @Computing_News / @Tom_CTG