Tuesday, October 1, 2019
At a time when AI is permeating its way throughout the Semiconductor industry the EU’s decision to protect personal data could be seen as a death blow to R&D in the region but CEA-Leti, the French technology research institute, thinks not.
Speaking to EETimes Asia at this year’s Semicon Taiwan, CEA-Leti CEO Emmanuel Sabonnadière spoke of the benefits of keeping data local and the role Edge AI will play in meeting important industrial, medical and societal challenges.
“Data protection isn’t killing our industry, it’s helping us re-invent our industry and Edge AI is an example of that,” Sabonnadière said.
The ‘real’ application for AI
CEA-Leti chose to "abandon" their efforts on Cloud-based AI and to instead put all their eggs in one basket with Edge AI. Sabonnadière believes that it’s healthy to focus CEA-Leti and that there is already plenty of work being done on cloud-based applications. To be clear, when CEA-Leti refer to ‘Edge AI’, they are talking about AI which is structurally integrated into the system to enable autonomy from the cloud.
Sabonnadière explains that CEA-Leti “strongly believe in this investment long-term,” and feel that once AI is ready, the “real” application will be locally. We’re starting to see the first applications around Edge AI with MCUs but, “in 3-5 years you’ll probably see the first NPU dedicated for some Edge AI applications and mainstream NPUs are probably 5-10 years out,” said Sabonnadière. Once we reach this stage, Edge AI is likely to be everywhere in our daily lives.
French startup Diabeloop, working in partnership with CEA-Leti, is a great example of how Edge AI can be utilised in the automated treatment of type 1 diabetes. Today, a diabetic is in a constant loop of blood tests, daily exertion analysis and having insulin injections. Diabeloop are currently using AI to locally learn the patient’s requirements so that it can personalise their treatment.
Currently, they are using AI to make 60% of the decisions with a goal of 95% in the future. “Edge AI needs to be able to understand everything and make all the decisions and then you’ll have the dream case of the AI controlling the loop all day long,” said Sabonnadière.
“It’s about the utilisation of your data with you, for you, which will make Edge AI strong,” said Sabonnadière. He believes there will be 1000’s of unique applications but wanted to highlight how Edge AI could help fight fake news.
Fake news has become a blight on society in recent years, no more so than around election time but Sabonnadière believes this is another area where the power of Edge AI will be felt. He envisions a future when the AI on our phones will know what we’re working on and what we like, possibly even guiding us towards more content that matches our interests.
This all seems very similar to browsing Facebook today: however, the AI on our phones will have the capability to review the validity of the articles and determine if it belongs in our personal network.
Is this a question of ethics?
Is this all just marketing spin or have Europe found a means of carving out their niche when faced with strict data privacy regulations?
Sabonnadière was quick to express that we’re still “in the middle of the game” and that it would be considered “very arrogant” to know what the future holds for the success of AI in Europe. But for Sabonnadière, this is more than just a question of meeting the demands of regulations, it is about ethics.
He believes that “for myself, my family and my friends, I do feel that this is the right thing. You cannot have your data floating around out there with people you don’t know taking care of it and developing things on the back of it,” said Sabonnadière.
It is now so easy to freely provide a company with our personal information, for most people they probably don’t even consider the intrinsic value their data holds. This status quo of uploading our lives to the cloud for unknown organisations to manage – in many cases without the protection of anonymity – is utterly baffling for Sabonnadière.
Sabonnadière argues that personal data which enables an organisation to draw a conclusion on an individual should stay within your own “personal network”. “If I chose to run a system to help identify a disease, the result has to be for me, the insurance doesn’t have to know that I’m sick or not,” Sabonnadière explains.
Copyright © 2019 CST, Inc. All Rights Reserved