This talk will explore future shifts in computing paradigms and the architectural implications for computing systems. These future systems must responsibly support the desire for extreme compute throughput and provide even tighter integration into all facets of life. As we race towards AMI (Autonomous Machine Intelligence - aka AGI), building and powering large datacenters that train and serve AI models, we risk burning up the planet. Emerging frameworks allow us to estimate the direct environmental impact of systems in the early concept phase and experiment with tradeoffs. Using these frameworks, computer architects will be able tradeoff environmental impact with power, performance, and cost. The encroachment of AI into nearly every facet of our lives will drive a paradigm shift: wearable devices will require the system capabilities of a laptop or smartphone, while maintaining a small form factor and low power consumption. The devices will also need to incorporate sophisticated, always-on, sensor systems that honor privacy and implicit social contracts. Eventually, these devices will be capable of replacing laptops and smartphones.

Biography

Doug Carmean is currently a Director of Research Science at Meta where he leads a group investigating future augmented reality systems. Previously, Doug was a Distinguished Engineer at Microsoft exploring the role of advanced technology in the context of future datacenter computing ecosystems. Doug has been deeply involved with the development of quantum computing, superconducting and classical computing systems. Doug led a group that investigated optics-based fabrics for future datacenter deployment. As an Intel Fellow, Doug was responsible for creating the vision and concept for the Xeon Phi (Larrabee)family products, an architecture for highly parallel workloads based on Intel Architecture processors.