Westworld — The Next Analytic Frontier
Storytelling, Feedback, Cognitive Science, and Deep Learning
**This article is NOT Spoiler-free, but contains as few as possible and few if any beyond episode 2.
The new Westworld series on HBO has been well received. It is a franchise nearly a half century in the making. Michael Crichton first released Westworld in 1973. The story line progressed through several more movies before dumping robots entirely and adding dinosaurs. In 2016, the original premise was mostly restored and with Anthony Hopkins as the lead, it launched again on HBO.
It is a story of iteration and development, on two fronts. It is the story of scientists playing god and of new species evolving and learning. But unlike Jurassic park, it is not the interaction of DNA strands. It is a tale of storytelling, feedback, cognitive science, and deep learning.
It begins with memories.
…or reveries. Hopkin’s character Ford is a scientist obsessed with telling stories. He believes in the subtleties and details. He seeks to expose the possibilities in his world. It is a world where robot “hosts” play the role of actors in a fantasy Western story line.
To facilitate this, he continues to tweak and improve his “hosts” in iterative fashion. Each of these iterations produces hosts that are more lifelike, more human. One of his techniques involves programming them to actually remember subtleties of previous existences, prior to this their memories were routinely wiped or reset.
Analysis
The series often employs the use of an “analysis mode”. The hosts are interrogated about their thoughts and motivations, by the park’s staff. These motivations are the driving force behind the personalities. These personalities are embedded in their code.
This includes sets of prime directives and other such rules-based constraints. Hosts are not allowed to hurt humans (new-comers). They are programmed to perform within constraints. Great care is taken to keep them there. But, of course, things don’t always go as expected.
Layers and Deep Learning
As the series progresses, the hosts begin to program themselves. It is a process a kin to the layers of deep learning. Hierarchies and goals form among the participants. Outcomes emerge outside of the expectations of much of the staff. The staff sees them as improvisations. The story line leaves the audience wondering about the reality of free will.
Memories, improvisation, feedback, and constant iteration become the very core of the story. The audience is provided a front row seat to learning and the emergence of consciousness. Or perhaps it is the illusion? The elements of cognitive science, while not perfect (not that the science even is) are portrayed in a surely thought provoking and compelling fashion. A story that iterated with Crichton for half a century, predicated on the deep learning of robot “hosts”, provides learning for the audience layered deep in an entertaining story!
Bicameral Man
Bicameralism was a philosophy published a few years after Crichton’s first movie and screenplay. I don’t know what the inspiration was to pull it into this story line, but it connects brilliantly. Bicameralism, which is not a well-received concept among scholars, claims that prior to about 5000 years ago, man had no free will. Our minds were split between transmitting and receiving.
The voice we were receiving was the voice of god. At some point, this voice was displaced by our own consciousness. It is fitting title for the final episode. One the writers play with nicely and manages to keep a certain subtlety hidden below other layers of plot twists. The philosophy engenders the same skepticism in our fictional tale that it gained in the real world.
Failure & Learning
This is a concept I have written on repeatedly. The series takes it further, replacing failure with suffering. It is still a strong lesson in the value of negative feedback, but from my own canon — it may be a magnitude too far. Or perhaps not…
Good, bad, improvised or scripted, free or controlled — this show will leave you guessing. It was a great 10–11 hours of content and I will have a front row seat for season 2. Unfortunately, we will have a bit of a wait. Thanks for reading!
