Be it a terror-filled paranoia or an anticipation-filled thrill, almost all of us geeks have some opinion about artificial intelligence and what it could do for us in the future. Disney, with NVIDIA, has come up with something that could give you a fair amount of reason to feel both at the same time. It’s a piece of tech that can monitor moviegoers as they watch the silver screen and gauge their reactions and facial expressions. This information could then help Disney make better movies and get a lot of additional information on what people feel about their movies.
The tech angle
Disney is putting into use NVIDIA GPUs as the star of this particular big screen act. With the help of deep learning, a particularly popular branch of AI right now, Disney’s AI tech is trained to observe the faces of moviegoers as they experience the movie in front of them – and track their smiles, frowns, wide-eyed fear and more. And it’s not just the overall mood we are talking about here – this tech can track reactions down to a very granular level which would enable it to provide much more accurate data than other methods such as interviews and questionnaires.
The actual method being put into use here is something known as factorized variational autoencoders (FVAE). Deep learning allows an AI system to process tonnes of data (the moviegoers’ reactions in this case) and make inferences from the data. NVIDIA’s Tesla K40 GPU’s are being used by Disney researchers for high-performance computing as well as accelerating the time it takes to train these neural networks and more.
There could be quite a few interesting applications of this technology if it is actually used. For starters, Disney could use it with test audiences to get a better picture of how their movies, or specific portions of the movies, would do when released and use the data for any last minute changes. Additionally, this also opens up the door to the idea of interactive or reactive storytelling, where the story of the movie changes based on the audience’s reactions.
Disney could also put this to use at any of its many theme parks to judge the reactions of visitors at specific rides. One big issue with all of this? Privacy. While that could be worked around with test audiences by providing a disclaimer and allowing them to not participate if they do not agree to be watched, a similar process for actual moviegoers and theme park visitors might just not work so well.
Even if this technology does not make it beyond the research stages, one thing is certain – GPU based deep learning applications in AI are going to be the way forward in the area, be it for driverless cars or for tracking your facial expressions at a cinema near you.