The new year’s first MadHATTERS op ed, must surely feature Irene Ng’s and Hamed Haddadi’s article in Wired Magazine’s 2019 predictions where HAT was featured as the “Edge AI” solution.
From the Wired 2019 edition, 28 December 2018
Most AI giants on the internet rely on the continuous collection of personal data from their users, primarily to build and maintain machine-learning models. These models are often core to the value proposition of these companies, providing recommendations, behavioural analytics and consumer insights not only to their own services, but to associated advertising networks.
This practice, however, comes at a cost to individuals. The repeated delivery of ads by third-party services creates excessive bandwidth and energy usage, something consumers are noticing as ongoing data collection and analysis by background apps slows their internet connection. And, as many recent cases have shown, there are now serious privacy concerns from excessive data collection and the resulting exposure from linkages of personal data across different services.
In 2019, we will see an alternative to these practices emerging in the form of AI at the edge – machine learning that will take place “near” the user, on their device or home hub, or at a local data-aggregation point. This will take different forms, including local learning (where the model is trained locally); distributed or federated learning approaches (where a globally trained model is optimised and retrained locally without transferring data back to the cloud); or co-operative learning approaches (where local data contributes to a global model on an ongoing basis).... Read the rest of the article here.
Yours in HAT,