Will Google bring deep learning to the deep learning to the device? It appears that will be the case. In fact, AI is expected to be deployed mostly at the edge, and not in traditional data centers.
Traditionally, hardware limitations have made this too prohibitive. But prices are really dropping.
Google CEO Sundar Pichai says Google deep learning models, which were previously up to 100 GB in size, have been scaled down to 0.5 GB.
“What if we could bring the AI that powers Assistant right onto your phone?” said Scott Huffman, VP of Engineering for Google Assistant.
For purposes of clarity, smartphones have not traditionally been considered “things.” That appellation has been reserved for other appliances or sensors not intended to be used by humans.
For purposes of plotting the value of edge computing, the distinction might not matter, as much of the consumer IoT space (smartwatches, for example) features devices which are used by humans.
No comments:
Post a Comment