The Uncanny Valley of the Mind
Back in the 1970s, as the first humanoid robots were developed, Japanese roboticist Masahiro Mori did his pioneering work on the emotional response of humans to non-human entities. He coined the term ‘the uncanny valley’ to explain the perceived likability of robots. It became one of the more successful viral ideas in the history of robotics.
Essentially, ‘the uncanny valley’ hypothesises that the likability of a robot increases the more human-like it gets, until it reaches a point beyond which our emotional response quickly becomes a feeling of weirdness, revulsion, or even fear. However, as a robot’s appearances are developed to near-full levels of realism — think HBO blockbuster TV-series Westworld — it will re-emerge from the valley and be met with more positive responses.
While Mori’s uncanny valley is concerned mainly with the physical and aesthetic appearance of robots, we might fall into a different, potentially much deeper uncanny valley with the rapid advances of artificial intelligence. An experimental study by psychologists at the Chemnitz University of Technology presents evidence for a much more concerning uncanny valley — ‘the uncanny valley of the mind’ — which relies on the attribution of emotions and social cognition to non-human entities.
The experiment found that people react significantly more negatively to avatars if they are presented to be artificial intelligences with the ability to make their own decisions and reactions. Avatars that were presented as human-controlled were perceived as far less unsettling, though their behaviour was identical to the allegedly AI avatars.
It is not AI as a technology that is off-putting to us. Today, we often interact with AIs in our daily life — also more often than many people realise. If you own a smartphone, chances are that you have been using Apple’s Siri or one of the other intelligent personal assistants out there. Yet, as long as Siri speaks with a computer voice and is distinctly not human, we don’t feel too threatened. You are probably also not overly freaked out by the fact that Netflix often knows your movie preferences better than you do yourself through its machine learning algorithms.
We might already be well past the classical (physical and aesthetic) uncanny valley. It is no longer the way robots look that will bother us, but how they act. Clearly, there are still many things humans can do that computers can’t. However, AI already rivals human intelligence in many aspects. The ‘brains’ of the most advanced computers and robots are able to do things that will make many of us shudder uncomfortably. As robots and computers start mimicking traits and abilities that we once thought were uniquely human, we begin to feel unease — traits such as social cognition, the ability to feel and understand emotions, and to show empathy. It simply threatens our human uniqueness, while also evoking our subconscious fears of reduction, replacement, and annihilation.
We might already be well past the classical (physical and aesthetic) uncanny valley. It is no longer the way robots look that will bother us, but how they act.
As developments in AI continue there is a good chance that we will experience human-level AI in our lifetime. Some technologists boldly predict that human-level AI is just around the corner — the most radical even say that it could happen as soon as 2030. We might be standing on the edge of a steep cliff looking down on a vast eeriness — we might even already be sliding down the slippery slopes of the uncanny valley of the mind.
