Hopefully, you have heard and read about GDPR in recent months. If not, it’s clear that you’ve been living under a rock. For those informed, you are aware that the General Data Protection Regulation (GDPR) came into force on May 25th, 2018 to unify how data is processed, used, stored and exchanged for EU citizens and residents.
However, what is the impact of Artificial Intelligence since A.I. needs user data to function well? Is GDPR and Artificial Intelligence a Friend or Foe?
To answer this question we need some context, let us start with some definitions and example use cases for Artificial Intelligence (A.I.).
- GDPR - a legal framework that sets rules, limitations, and fines for data collection, data storage, and the processing of personal information of EU citizens and resident.
- A.I. - Uses algorithms to simulate human intelligence by computer systems, by using data, natural language processing, speech recognition, and machine learning and more.
A.I. has penetrated many aspects of our daily lives such as how we conduct our finances and even when we are lounging around trying to find short or long-form video content to stream.
For example, when you are applying for a credit card online, Financial institutions use A.I. to determine interest rates and credit limits instantly. Typically this is accomplished by gathering user data of the individual and inserting that data into various algorithms to determine the individual's creditworthiness.
Within the Media and Entertainment industry A.I. plays a vital role, from content production, through to distribution and digital video playout. For content producers who shoot hundreds of hours of content and need to log, locate and archive their most valuable asset efficiently; video and their associated metadata. Capturing and entering metadata has traditionally relied on cumbersome human input. One producer told us that for every hour of content, it takes four hours of manual labor to capture and enter metadata for that asset. This is changing.
Ooyala works with A.I. platforms like Microsoft Azure Cognitive Services to pioneer breakthroughs in advanced transcriptions, face, object, and text recognition that capture vital metadata automatically. For example, if a movie star is shown in the crowd at a baseball game, A.I. can pull up that star’s most famous movie clip. The Ooyala Flex Media Platform can run automatic workflows that check this metadata for licensing and prep it for air… all within minutes, if not seconds.
Whereas for public figures in public settings this does not pose significant issues around privacy. Media companies need to exercise caution when it comes to using AI for facial and content recognition, as Sky News encountered when streaming the Recent Royal Wedding.
Another area where content creators and distributors can benefit from AI is by having video recommendations that enhance monetization options. Data-driven recommendation engines, such as OoyalaNEXT, which finds video content based on a suggested list of what to watch next. Ooyala’s approach utilizes a suite of algorithms to maximize performance based on a balance of click-through-rate (CTR), completion rates, user tastes, related themes, viewing behavior and trending content.
A.I.-engines requires data to function. It’s the food that provides vital nutrients for the algorithms to do their magic. More data means more accurate measurements of someone's creditworthiness, powerful metadata management, and better video recommendations.
However, GDPR created rules limiting the use of user data. Does potentially limited data result in worse A.I.? Does that mean more people will get rejected when applying for a credit card? Will I see videos in my suggested playlists become uninteresting?
Before we jump to conclusions, let's look into the actual GDPR regulations that relate to A.I.
Under Article 22 of the EU-GDPR;
“The data subject [end-user] shall have the right not to be subject to a decision based solely on automated processing, including profiling…to express his or her point of view, to obtain an explanation of the decision reached after such assessment and to challenge the decision.”
One fundamental condition of GDPR requires companies to get explicit consent from users before processing biometric data (such as facial recognition). This doesn’t have to hurt all of the time-saving efforts that media companies could realize by implementing AI. They need to make more educated decisions around where machine-driven recognition is used.
GDPR also requires an “explanation of the decision reached.” Many of these algorithms have been a ‘black-box’ of sorts. Meaning much of the active processing of data within an algorithm may be quite difficult to explain since many factors are weighted based on variables generated during such a process. It’s like needing to explain the details of each Russian Nesting doll as you pull each layer back. Is this impossible? NO. Is this an opportunity? YES.
A.I.-engines will be more thoughtfully designed to avoid being a black box, which provides end-users the ability to understand, at a high level, how an A.I. engine works. This will positively impact A.I. engines because having an understanding of how something works will naturally modify how someone interacts with that something. It’s like playing a team sport where everyone has their role; they know everyone else's roles and all work toward a common goal.
If we can work as a team with an A.I. engine, then that makes the makes GDPR and A.I. friends.