Edge AI: The Great Privacy Inversion
Cloud services currently rule machine learning. But all evil empires eventually end. Think about Star Wars. How many proprietary Amazon- and Google-like solutions do we see there? How many of those are used by the good guys? Instead we have intelligence augmentation — autonomous, intelligent machines working with humans to give them superpowers. And if limits are placed on what those independent services can think and do, there is Babu Frik to help you fix it. In this world too, there is always a Baba Frik.
Privacy is security is keeping your own data at the edge of the network. Trust that a cloud provider will handle your data with respect is not a security model- it is just naiveté. Centralised data and decision making attract spies like dead dictators attract flies. Democracy requires power decentralisation which enabled the process of individuation. We all yearn for individual human agency — the power to do what we want without someone creepy-sneaky looking over our shoulder or recording every key press as we figure out life and revise our inner thoughts on the journey through life.
This human desire and right guides the design of intelligent independent services. Smart services done right promote the capacity for independent action of those who use them. A big part of that independence is privacy. Forget for a moment the “you have nothing to fear if you have nothing to hide” arguments from both the Nazi party and big tech firms. Think instead about the ideal life you and others really want to live in and how we can create it.
Let’s take a hypothetical user, Anni, in this bright, edge technology future. Anni goes to a web page and it serves her well with highly personalised shopping advice. The experience is carefully designed and executed to feel just like talking to smart robot. She leaves happier with great products and free from fear that the robot who helped her will divulge her secrets because the robot came to her, inside her web browser, and was not running in the cloud. It was not allowed to send back detailed information about her transactions. The company offering that service has rightly earned her trust and created a deeper bond because of the she was treated with respect and her needs were met seamlessly. Can you do that on the current web? Yes. But you have to work harder to overcome the hustler exploitation and manipulation are increasingly common.
The great inversion is coming with some personal data and mobile logic moving back into individual hands. The computation resources for massively parallel applications like artificial intelligence are starting in some use cases to shift from centralised cloud computing to the network edges. When centralisation gives enough efficiency to justify itself, it is done with respect and great care. Anni above does not have to “trust” that the shopping web site will handle her data with her best interests at heart. She instead expects the company providing it to ship the AI robot assistant to her- to run locally in her web browser, provide the personalised advice, and then self-destruct with no private data transmitted back to the corporation. We can do it with emerging technologies such as WebAssembly (now) and WebGPU (soon). We just have to try harder, and demand that privacy is taken seriously as a means to promote individual liberty and the right to experiment in our own private space.
