More than 200 million homes now have a smart speaker, according to a global estimate, providing voice-controlled access to the internet. Add this to the virtual voice assistance installed on a lot of smartphones, not to mention kitchen appliances and cars, and that’s a lot of Alexas and Siris.
Since speaking is a basic part of human being, it is tempting to think that these voice assistants should be programmed to speak and act like us. While this would give us a connected way of interacting with our phones, it is extremely difficult to replicate truly realistic human conversations. Moreover, research suggests that it may be unnecessary and even dishonest to make a machine sound human. Instead, we
Recent developments in the development of artificial speech have resulted in the voices of these systems blurring the line between human and machine, sounding ever more human.
Google introduced utterances of voice assistance such as “hmm” and “uh” to the speech performance of its assistant to the human-like nature of the system–sounds which we typically use to indicate that we listen to the conversation or that we intend to start pace.
Chasing this goal of making systems sound and behaving like us that derive from the inspirations of pop culture that we use to drive the development of these systems. For decades, the idea of talking to computers, through characters, has fascinated us in literature, television and film. In trying to achieve something that resembles conversations between us and machines, there are interesting technical challenges.
These great challenges are clearly advancing the state of the art, like others across science, bringing planned and unplanned benefits. However, as we strive to give machines the ability to really talk to us like other human beings, we need to think about what our spoken interactions with people are for and whether this is the same as the kind of conversation we want with machines.
Looking beyond humanness
Instead of constantly promoting humanity, we should recognize that there may be inherent limitations, both technical and theoretical, to the types of experiences with machines that we can and want to have.