On-Device AI and Data Privacy Concern

The amount of our private data being shared through the usage of tech-based devices, mobile applications, and other cloud-based services is and sometimes even without our knowledge in the least understood aspect of it. However, we’ve all started to become painfully aware of how big (and far-reaching) the problem of data privacy is over the past year. Owing to this, there has been an enormous spotlight placed on data handling practices employed by tech companies.   

Simultaneously, the expectations about technology’s ability to personalize these apps and services to meet our specific interests, location, and context have also continued to grow. People expect technology to be “smarter” about them as it makes the process of using these devices and services faster, efficient, and more compelling.

Enabling this customization requires the use of and access to some level of personal data, usage patterns, etc. and that’s where the dilemma lies. That has typically meant, up until now, that most any action you take or information you share has been uploaded to some type of cloud-based service, compiled and compared to data from other people, and then used to generate a response that’s sent back down to you. This gives you the kind of customized and personalized experience you want theoretically at the cost of your data being shared with a whole host of different companies.

Thanks to the AI-based software and hardware capabilities becoming available on our personal devices, more of the data analysis work could start being done directly on devices starting in 2019, without the need to share all of it externally. The idea of doing on-device AI inferencing is now becoming a practical reality thanks to work by semiconductor-related companies like Qualcomm, Arm, Intel, Apple, and many others. This means that if app and cloud service providers enable it, one could start getting the same level of customization and personalization they’ve become accustomed to, but without having to share their data with the cloud. However, some of your data will still be shared inevitably because it isn’t likely that everyone on the web is going to start doing this all at once (if they do it at all). If some of the biggest software and cloud service providers like Facebook, Google, Twitter, Yelp, etc. start enabling this, it could start to meaningfully address the legitimate data privacy concerns that have been raised over the last year or so.

Apple started talking about this concept several years back through differential privacy and already stores things like facial recognition scans and other personally identifiable information only on individuals’ devices. We can expect to see many more hardware and component makers take this to the next level over the next year by talking not just about their on-device data security features, but also about how onboard AI can enhance privacy. Let’s hope that more software and cloud-service providers enable it as well.