AI as Mind-Reading
A few days ago, a fleeting memory crossed my mind: an old Warner Brothers Bugs Bunny cartoon, where our long-eared and cotton-tailed hero somehow winds up in Little Red Riding Hood’s picnic basket. Little Red, departing from her usual characterization, is a loud, obnoxious boor. Thoughts like these occur all the time to human beings, and I thought little of it, my stream of consciousness drifting off to other topics and other images.
Fast-forward to yesterday. I open YouTube and, to my amazement, on the top of my feed is a clip from that very cartoon:
https://www.youtube.com/watch?v=arDVQ-C44tM
I was astonished. Can AI read minds now?
Upon reflection, I calmed down, and, in Socratic fashion, I interrogated my assumptions and tried to integrate them with what I knew about generative, neural net AI. Here’s what I came up with.
AIs gather data constantly from one’s web surfing, online transactions, and even Siri or Alexa listening in. At some point, these data were analyzed by multiple layers of “Deep Learning” style AIs (DLAIs) on YouTube, which purchased access to these data. Remember, since DLAIs have different statistical processing layers, the software engineers often do not know what statistical connections could be made perusing my browsing habits: they do not know how the DLAI learned how to recognize…
