Practical application of social data streams is something that should change a LOT to do with our lives. Here’s one example, with all sorts of interesting ramifications…your mood and reception to the movie you’re watching.
I saw this post about some work Disney is doing to figure out how people respond to movies. I think it would be fascinating to see the information they’re capturing and how it could be used to … make movies better?
First, here’s a link to the article: https://www.fastcodesign.com/90134144/disneys-next-movie-it-could-be-watching-you
I’ll admit though…I’ve thought about this and don’t know how I feel about formulaic movies that illicit specific responses. At first, I think it could be gratifying or entertaining or whatever you want to call it to be able to pick a movie that might impact you a specific way, and have it do that. But a few things come into play with that.
First, you’re picking – you’re picking the expected responses. Second, it’s so… cold? Planned? Controlled? I mean, I’m trying to decide if you would pick a war movie if you were “guaranteed to be repulsed by the gore of the battle scenes” and all that.
It seems, too, like it could remove the edgy movies – the ones that impact different people different ways.
But that’s not actually the point of this post (believe it or not). It’s more about data – and processing the flows. I’d love to see what data elements are being tracked. In the linked article, you can see the image that they use for the post at least has an emphasis on fairly specific points on a face. I don’t know if those are actually what they track, but I was trying to imaging making sense of the data elements across an entire audience so you could get a good grip on what was working.
Then they mention that something like 65M data points were captured. That stream of data (bringing it back to SQL Server and databases) is substantial. If I’ve done the math correctly, that’s a lot of information flowing – and catching differences in a meaningful way for different facial features (beards, smirks, embarrassment, etc.) would be a huge understaking. I realize they’re not processing in real time, but…
On the way to work today I also was listening to a story about the work they’re doing with autonomous cars. Trying to determine the data points associated with bike riders and how they influence and respond to traffic and trying to get predictive about where they’re going before they go there. They’re learning these things by a whole lot of sensors on bike riders, looking to determine what the critical elements are and how to interpret them in very real (life-saving) time.
I suspect this is one of the first really significant payoff milestones of AI and big data. To be able to read humans and know more about them, to “understand” and to be able to respond and interact and react to someone based on what’s known about actions and reactions from people. The little inferences pulled from massive data sets start to get really interesting when the data set is huge – but at the same time, doesn’t the specific learning get dilluted if the data set gets really, really big? Is there some cross-over point where the database starts to generalize itself away from usefulness?
Inquiring minds want to know.