News of the Google AI assistant (or Google Duplex) being used to place phone calls to people in such a way that the human would not know they are speaking with AI was a little unsettling, and Google has announced they will address this by having the AI-driven, interactive robocaller (for lack of a better word, if you know one please suggest it!) introduce "itself" as such.

My question addresses this or similar scenarios when one is confronted by human-like AI presenting cues that it is artificial, and yet including verbal affections that are specifically intended to sound human (e.g. 'um, okay...') rather than a machine.

The concept of the uncanny valley is usually applied to visual stimuli, but considering the degree to which AI and realistic human-like speech can be readily generated and transmitted, I'm wondering if the uncanny valley concept could apply or be extended to aural stimuli as well. Also curious if this avenue has ever been explored in published research.

One problem with my question might be in calling the uncanny valley a concept, rather than a phenomenon associated with visual stimuli by definition. If that's the proper way to think about it, perhaps I should be asking if there is an aural analog to the uncanny valley instead?

  • 2
    $\begingroup$ I don’t think it changes the question itself whether it’s an aural manifestation of the concept called the uncanny valley or if it’s the same experience with a different name. Then both the uncanny valley phenomenon and the un-named aural one would be two different manifestations of the larger thing you’re asking about. The discomfort with the too-nearly/not-enough human is pervasive isn’t it? An AI that stammered would be off-putting, because it’s obviously manipulative. An AI doesn’t require appearing or sounding human for any purpose intrinsic to its own intelligence. $\endgroup$
    – Jeannie
    Jun 17, 2018 at 18:14
  • 1
    $\begingroup$ Note that Wikipedia has a long Criticism section for Uncanny Valley, and to my knowledge, there is no compelling evidence that the effect is real, popular as it may be in the media. For more details see: The Uncanny Valley Is Wrong. $\endgroup$
    – Arnon Weinberg
    Feb 11, 2021 at 23:42
  • $\begingroup$ @ArnonWeinberg thank you for your comment! It looks like my question premises that the phenomenon is real, but as your comment and the article suggest, it's more of a well-worn hypothesis. Of course any good Wikipedia article about a hypothesized phenomenon would include a criticism section, that's what hypotheses are for. (It can't be a hypothesis unless it has the possibility of being disproved.) My nonprofessional guess is that if people were asked if they ever experience feeling "creeped out" by near realistic representations that are just slightly "off" somehow, most people would say yes. $\endgroup$
    – uhoh
    Feb 12, 2021 at 1:31
  • 1
    $\begingroup$ @ArnonWeinberg and so, perhaps, each person has their own uncanny valley, and it tends to get washed out in standardized statistical studies that look for common responses or reactions. Hmm... Anyway Wikipedia linked to io9 which is taken from The BBC's Robots: Is the uncanny valley real? Thanks for giving me something to think about today, please feel free to write an answer calling my premise itself into question. $\endgroup$
    – uhoh
    Feb 12, 2021 at 1:35


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Browse other questions tagged or ask your own question.