MACHINE LEARNING
What is TinyML ?
Additionally , according to Noury , TinyML is becoming increasingly relevant as manufacturers are investing more into edge computing .
“ Edge networks help with the surge in performing AI tasks at the edge , which require distinct capabilities that help to handle elastic demands and ever-changing workloads ,” he said .
On the other side of the physical world spectrum , according to tinyML Chairman Gousev , sensors are becoming more sophisticated , sensing a variety of modalities such as vision , sound , environmental and motion / vibrations .
“ These are being deployed in their millions every day . The “ collision ” of these two powerful driving forces , tinyML and tiny smart sensors , establishes a unique opportunity for edge computing and on-device analytics that will be happening everywhere and anytime . As an end result , this will create a better , healthier and more sustainable environment for all .”
The emerging area of running AI models in large fleets of small IoT devices Edge AI is a part of the TinyML space , addressing a broad category of use cases and target devices , according to Tomas Uppgård , Head Of Partnerships at Stream Analyze , a Swedish company helping large industrials to create Edge AI solutions .
“ More and more data is produced outside of the cloud directly in IoT devices and other equipment with in-built sensors or computing capabilities ”, says Uppgård .
According to Gartner , 75 % of all data produced will be in devices outside of the cloud by 2025 . Uppgård says it isn ’ t possible or even desirable to send all this data to the cloud for analysis : “ The concept of Edge AI is instead about sending the AI models to the devices for analysis where the data is produced . It is a key enabler to utilise the data produced for intelligence in connected devices ,” he said .
124 April 2022