AI At The Edge Is The Next Major Technology Shift

AI is emerging as one of the most important technologies we need to track. From managing our homes and autonomous cars to detecting and reacting to cyber threats, making decisions about where to spend money or even determine prison sentences, AI is the pivotal technology. Dr Chang Huang, the founder of Horizon Robotics kicked of day of CES Asia saying that AI applications and edge computing are combining to create a new model for how tech is deployed.

While 5G and cloud computing have dominated headlines over recent year years, they have a significant challenge. No matter how fast the cloud servers are and the 5G can transmit and receive, latency is still a major issue. To put that in some context, Huang said that 2000 autonomous cars generate as much data as all humanity did in 2015. There’s simply no way split second decisions and actions can be executed on the cloud.

What is required, said Huang, is a high availability platform with low latency and that can comply with data privacy and security regulations. Edge computing devices. can fit the bill.

For example, in a car, a processing unit can receive the data from sensors, process it and complete an actions without relying on a data network. Data that is useful to the collective can be redacted at the edge and sent to the cloud at a time when latency is not a concern. This is also potentially less expensive than the cloud.

The cost, both financial and environmental, of cloud computing is something the IT industry will need to contend with soon said Huang. With cloud computing already producing more greenhouse emissions than the airline industry, low cost and low power edge devices can reduce the load on cloud data centres. Huang says there’s a social responsibility on developers to ensure they use resources as efficiently as possible.

There is a trade-off though said Huang. Cloud services are useful when there are lots of different data inputs and there’s a need for lots of computing power. For example, when creating new AI processing models. But edge computers are optimised for specific scenarios. In a home, an edge computer could be used to control a climate control system. But if an unseasonal weather change comes, then a new model could be developed on the cloud and sent to the edge device.

Those who gave been in the IT game for a while understand the elasticity of various trends. We went from mainframes and time-sharing to client-server. Then we shifted to the cloud and subscriptions (sounds a lot like centralised computing and pay-as-you-use, or time-sharing, to me). Now we’re pulling back to hybrid IT and edge computing where more processing power is at the point of use.