This story contains spoilers for season 1 of Westworld, and episode 1 of season 2.
Westworld is back, and I’m more confused than ever.
Watching HBO’s robot drama is a little different from reporting on artificial intelligence, my job here at Quartz. Usually I’m trying to figure out whether some new AI breakthrough is BS, rather than trying to decipher how a fictional technology could make some kind of sense. While some things in the show, like how the robots can see and speak, are believable based on the direction that real-world AI is taking… I still have questions. Some may be answered in the season ahead.
What is that brain goo?
After the carnage at the end of season 1, Bernard (Jeffrey Wright) finds himself hiding from the rioting hosts. He tries to save an innocent host from the humans, but is knocked over, and leaks some kind of robot brain goo from his ears. The brain goo is also seen when a Delos security person extracts memories from a dead host’s head. The robot’s central processing unit (CPU) seems to be submerged in the stuff.
What is this brain goo and what does it do? Is Bernard’s brain overheating, and it’s a coolant? Does it insulate his CPU from some kind of radio wave on the island? Is it an energy source? (That last one seems unlikely, as hosts apparently function organically, both eating and pooping.)
Can the hosts be controlled remotely?
It makes sense that the hosts aren’t connected to the internet as we know it. You don’t want someone hacking a robot and stealing its code or its memories—not when guests are promised total secrecy. But there’s an ambiguity that irks me around what aspects of the hosts can be controlled over the park’s network.
We do see examples of remote control: Sometimes when human technicians enter a space, hundreds of hosts—near and far—will freeze. Is this an example of the host mesh network, as referenced at one point by Bernard? (A mesh network is a concept that dates back to the 1970s in Hawaii, where two computers far away can send information back and forth if there are intermediary computers to relay the message. If there are computers A, B, and C, in California, Dallas, and New York, then A messages B, which passes the message to C.)
If they can be controlled this way, why do the park maintenance people have to shlep out into the park and pick up the bodies? Why not just make them walk back home to get serviced? And if they can’t be remotely accessed, how do park personnel have such detailed information in the control room about which host is where? Why would they only broadcast their location?
Why is there no big red button?
Modern AI scientists are thinking about a kill switch—a failsafe option to kill an AI’s processes no matter what. Why doesn’t that exist in this world? And why wouldn’t that be automated if the host deviates a certain amount from its script?
Whose DNA do these robots have?
When Bernard grasps the DNA-reading handle of the secret Delos bunker, it reads his DNA and lets him in. When the showrunners said that hosts are “basically organic” and have been grown to have a CPU instead of a brain, then whose DNA does Bernard have, and is it different from the other hosts’? Wouldn’t the bunker recognize that he has host DNA?
What is the limit of the hosts’ intelligence?
Surely there has to be some kind of limit on how much data each host can compute—after all, they’re robots that don’t seem to connect to a processing cloud. (Any kind of cloud would mean the data would be accessible somewhere other than the robots, and wouldn’t necessitate Delos personnel manually extracting video.)
So what’s the limit to the processing power of their hardware? Can every host think 10 steps ahead of every human? Are there kinds of problems that the hosts can’t solve because they were only intended for limited use in a theme park? Is their intelligence really as broad and general as it seems? And if so, why?