The codependency of AI and the Edge

from my keynote at Edge Computing World: Couple some local intelligence to each sensor and the internet of things is becoming the sensory cortex of the planet, with countless data-collecting-devices. All of this ‘big data’ would be a big headache but for machine learning to find patterns to make it actionable, and edge computing to shift the processing to the periphery and avoid network overload. In short, the edge needs AI, and AI needs the edge. The compute architecture for machine intelligence is shifting to specialized processors optimized to the task, a biomimicry of the human cortex in compute substrate and algorithms. We are also learning how to improve the process of training with model compression and continuous learning.

The march to specialized silicon, from CPU to GPU to FPGA to ASIC is now going further, to analog and quantum processing. At a high level, we are recapitulating our evolutionary computational march in silicon, and an ever-growing percentage of our compute will be massively parallel and in-memory processing, just like our cortex.

More generally, progress comes from the edge — the edge of survival for evolution, the edge of energy gradients for life, the edge between chaos and determinacy for emergent complexity, the network edge for the sensory cortex of AI.

And in bullets:

Biological Muse for AI
• Modeled on the Brain
– Synaptic network at scale
– Iterative algorithm
– Transcendent
– Distributed sensory cortex
– Low-power, in-memory, analog compute (Mythic)

• Sparse Activations
– 10x Compression (Latent)

• Neural Plasticity
– Continuous Learning (Brdg.AI)

• AI needs the Edge => Sensory Cortex for the Planet

• The Edge needs AI:
– Intelligence dominates product features
– Better Latency, voice interfaces
– Robust to connectivity, off-line use
– Privacy, local data not shared to the cloud

With my neural network textbook from 1988. Photo by Rich Miller. Thx!