What Is A Robot?
Six years ago, at a tech conference in Dubai, I got into a spirited debate about what defines a robot. We were showcasing Gala, our first self-driving store at Robomart, and were sharing the show floor with other robots from across the world. In the booth right next to us, was Sophia the android with human like skin. After ‘she’ walked over and tried hailing the robomart, the founders present got into another argument on whether Sophia could be called a humanoid, or if that word exclusively applied to mechanical androids in the shape of a human, like the Boston Dynamic’s Atlas robot or Honda’s Asimo.
Those discussions kicked off a multi-year project in which I set out to define the true nature of robots, come up with a better classification structure and taxonomy for all robots past, present and future, and explore the role of robots in a future of non-human superintelligences.
That project has now culminated in Systema Robotica, my treatise on the order and evolution of robot-kind. The treatise seeks to better define what a robot truly is and find answers to the following questions. Do robots deserve to be defined by biological markers? How can we differentiate between a mechanical android versus one with synthetic skin? Will robots be our servants, equals or superiors? Will we merge with them to become human-robot hybrids? When robots become superintelligent will they be regarded as sentient? How can we even define sentience? What is a robot?
Defining a robot is not an easy task. If you ask ten different people you’ll get ten different answers. It will also vary depending on whether you ask a roboticist, a science fiction author, or a member of the public.
My definition of a robot hinges on its ability to be encompassing, yet clear in its scope, and one that considers a robot’s origin, design, and future capabilities.
A robot is an artificial material construct designed to autonomously sense, decide, and operate within the physical world.
This definition helps not only specify what constitutes a robot, but critically allows us to consider what a robot is not, by comparing it to other things, such as machines, artificial intelligence, humans and cyborgs.
Neuralink has recently shared efforts in implanting their first two human patients, who can control physical objects simply by thought. This level of virtual-physical manipulation will eventually become more like a form of telekinesis through advanced brain-computer interfaces, unlocking new senses beyond the five basic ones humans experience. The trend of augmentation will only grow over time, further blurring the lines between what it means to be a robot, human, or cyborg, and how each of these will evolve in the future.
To make better sense of this, we need to consider forms of intelligences as a separate realm, that sit above the animal kingdom and biological domains. There are three theorized realms of intelligences that have the capacity to evolve: biotica, all natural biological intelligences, robotica, all constructed artificial intelligences, and exotica, all hybrid exotic intelligences. With the top level hierarchy defined, we can then start to codify robots past, present and future into a taxonomy under the realm of robotica.
The vast majority of taxonomies for robots to date have classified robots based on functionality. This is a fool’s errand, in that there will be considerable overlap making any taxonomy completely unusable. Androids designed for the kitchen can just as easily work in a factory, and drones built for agriculture can easily be reprogrammed for aerial photography. Instead, by focusing on a robot’s “designform” — its design, form, countenance, size, shape and branding, we can create a taxonomy that works for all robots.
If we were to create a taxonomy of humans, we wouldn’t classify them by functionality. Humans can do so much, hold so many roles, but we all have a specific visual countenance and appearance. The logical approach is to classify based on design and form rather than functionality.
The Robot Taxonomy defines six major types that all robots fall under. Within those six types are 22 schemes, based on a robot’s schematic designform.
The schemes span not only traditional robots that we’re familiar with, like androids with synthetic skin (synthoids), factory arm robots (articulators), self-driving vehicle robots (autons), and flying robots (drones), but also considers future robots and edge cases, such as robots that can morph, or change their appearance, countenance, size or form. MIT’s M-Blocks is one example of a shape-shifting robot. These robots are classified as morphbots, under spectra, one of the six major robot types.
Under spectra are also virtuoids. These are virtual or holographic robot entities, within a physical housing. Since they have an embodiment and meet the definition above, they are robots. These virtuoids are not science fiction but rather a reality. A company called Gatebox out of Japan has created a robot that projects an anime character imbued with artificial intelligence, that can talk to you, manage tasks for you, and even welcome you when you get home.
Another example of a virtuoid (this time from science fiction) would be Joi from the movie Blade Runner 2049, which was released a year after Gatebox introduced their virtuoid.
Joi is a holographic representation of an artificial intelligence, yet one that has a physical presence and embodiment through projectors that can interface with the physical world, and because of that, it is a robot.
As more and more startups get into robotics, we now need a clear naming system within the taxonomy. Joi would be the robot’s model name, as there can be dozens or hundreds or thousands of different individual unique robot units that are all the same model. Similarly when we think about a robomart self-driving store, our first robot is named Gala. This refers to the specific individual robot unit within our first generation model.
The impact of robots on society will be unlike anything we’ve experienced as a species. Robots will be able to hold one of nine distinct roles in human society, based on their degree of intelligence and autonomy, and their strength of relationship and emotional bond with humans.
Most robots today fall under robot as tool, with startups pushing to make robots more useful, autonomous and intelligent. There is a possibility, however, that once robots attain a high degree of autonomy and intelligence, if we disregard their complex natures, and only interact with them as mere tools they will be the equivalent of slaves. Over time, roles will move across the matrix, and as humans become exceedingly intimate with robots, they will start treating them as life partners, companions, and even going so far as to build a life exclusively with them.
This may feel far fetched today, but consider that intelligence is growing at an exponential rate.
Many believe that superintelligence in robots will lead to an existential crisis. However, this is quite an unlikely scenario, as there is no basis to believe that robots will share the same feelings, desires, and motivations as humans, just by virtue of being superintelligent. Superintelligence does not equate to sentience. For robots to be considered a potential threat to humanity, they would not only have to be superintelligent, but they would have to be considered sentient, with their own feelings, motivations and desires.
This is why it is so important that we define sentience, as it relates to artificial intelligences. Sentience is a very hard term to define. Some regard it as having feelings or the ability to suffer, while others equate it to consciousness.
There are many theories on what constitutes consciousness, yet it is virtually impossible for a conscious being to experience another’s subjective experience. Until we have a definitive shared understanding of consciousness, we need to recognize that theories of consciousness are just that — theories, and therefore need to explore other, more straightforward means of determining sentience.
A simple way to visualize sentience is to consider a robot vacuum cleaner autonomously following a path to clean a living room. It could take breaks or change paths of its own volition, but we still wouldn’t consider it to be sentient. However if it pauses from its work to take in and enjoy the view and communicates as much to us, or asks a question unprompted that demonstrates curiosity, then a lot of us would assume it to have a certain degree of sentience. The challenge is that this can still be programmed or faked and won’t conclusively demonstrate sentience.
Therefore, in the treatise I have proposed a more straightforward conceptual framework to better define and clarify the concept of sentience, and on the back of that, a novel test for determining whether a robot is sentient.
In 1863, Samuel Butler published “Darwin among the Machines,” an article suggesting that machines might eventually become conscious and supplant humans. In it he introduced the idea of classifying robots but considered the undertaking beyond his capabilities:
“We regret deeply that our knowledge both of natural history and of machinery is too small to enable us to undertake the gigantic task of classifying machines into the genera and sub-genera, species, varieties and sub-varieties, and so forth…We can only point out this field for investigation…”
Over a century and a half later, I have endeavored to undertake this critical task. It has never been more important and pertinent, as a precursor to the automation age that is upon us.
Ali Ahmed is a roboticist, inventor of the self-driving store, co-founder & CEO of Robomart, mentor to Singularity University, robotics expert to Wefunder, and author of Systema Robotica, a treatise on the order and evolution of robotkind.
Systema Robotica will be published on August 16 on Amazon, and simultaneously released as an open access public work.
Visit systemarobotica.com to learn more.
