Are Spaceships like HAL Robots or Just Really Comprehensive AI Systems?

I was fortunate enough to catch the Envisioning 2001: Stanley Kubrik's Space Odyssey exhibit at the Museum of the Moving Image in New York. It is a surprisingly informative and interesting collection of the props made for the movie along with a collection of letters and production notes by Kubrick and Arthur C. Clarke. I highly recommend it and it really is hard to believe that the movie is over 50 years old- you can read my assessment for Science Robotics of the legacy of 2001 here.

The real star of 2001: A Space Odyssey is HAL, which some people consider just a computer and some, myself included, consider a robot. The debate shows up from time to time, for me most recently at last year’s We Robot conference where my co-panelist disagreed with my classification of Sidra in Becky Chamber’s A Closed and Common Orbit as a robot (see the RTSF review here).

So, let me offer my reasoning why HAL, Sidra, the Shadow’s Child, and the godzillion of other sentient spaceships are robots. A robot is a physically situated intelligent agent, as defined in my textbook Introduction to AI Robotics. An agent is an entity that can sense and act on its world. An intelligent agent is one that attempts to maximize its chances of success (within its capabilities). “Physically situated" means that the agent can sense and act on the external physical, not virtual, world. Back in the 1990s, artificial agents were either software agents (sometimes called softbots or knowbots) or physically situated agents (called robots), reflecting the emerging use of agents on the web versus in the real world. Even then that software/hardware dichotomy was a bit of an oversimplification as a robot was likely to share many of the algorithms and organizational principles with software agents (indeed, software agents actually came into being after the behavioral robotics movement led by Rod Brooks). The 2004 I, Robot movie expected moviegoers to be surprised that VIKI, the evil centralized computer, was not just a computer but gasp! could control things in the external world (like all the NS-5 robots). Perhaps moviegoers were surprised, but those of us in AI were not. The difference between VIKI and a NS-5 was pretty minor.

HAL is definitely a physically situated intelligent agent, so it is a robot. It could sense and act upon the real world and, indeed, could have conducted Discovery’s mission by itself according to the book. The whole of Discovery and its pods were its body- sensors and effectors.

I think the resistance to HAL being a robot is similar to some people’s resistance to self-driving cars being considered robots: if humans are inside, we don’t think of it as a robot. We tend to view robots are separate, like dogs or drones or Ava in Ex Machina. (The exception is giant pilot mecha like Pacific Rim or the Themis Files but whatever, see my take on that for Science Robotics here.)

Watch 2001 again and see if you think HAL is a robot!