SIGNALS

"So God created man in His own image; in the image of God He created him; male and female He created them." [Genesis 1:27]

I've always wondered why the popular voice assistants (Siri, Alexa, Google Assistant, Cortana, I'm looking at you) all default to a woman's voice. I'm sure that AI voice researchers have their reasons, and whether those reasons are sexist or not, I have no idea. But, the fact remains that their creations are definitely female.

Occasionally, you hear guys. For example, later this year Google will let you use Google Assistant with John Legend's voice. And, you can delve into the settings of the voice assistants, and change the voice to male. Also, various navigation systems let you pick and choose male or female voices, even with different accents. But, as far as I know, these are all binary decisions: female or male.

Well and good, but today the question of gender is far from binary. Forgive me for not elaborating, but those issues are far outside my ken. Suffice it to say that there is room in the cloud for voices that are neither male or female. There is at least a third option: genderless.

Virtue Nordic and Copenhagen Pride developed an AI voice named "Q." It is said to be the world's first genderless voice for artificial intelligence systems. Q's voice originated from recordings of the voices of real people who are nonbinary. They were combined to create what is judged to be a gender-neutral voice. You can listen to Q here. Q's voice is very interesting, at least to me, because it is something I have never heard before. While other AI voices strive to sound either male or female, this one does not. Q does sound very European, which is not surprising, given Q's European origins.

Now, I am probably entering a gender minefield here, but at least to my ears, Q also sounds female. We know that great care was taken to avoid a specifically male or female gender. For example, when developing the voice, prominence of fundamental frequencies was given to the 145 - 175 Hertz range, which is logical because it splits the difference between typically lower male voice frequencies and typically higher female voice frequencies. But, I hear female. It reminds me of the viral blue-black/gold-white visual illusion and the Yanny/Laurel aural illusion. For some reason, my brain wants to arbitrarily assign gender to the voice.

If you want to further explore the question of frequency, listen to Q again, this time at the developer's web page. The voice is centered at 153 Hz, but you can move an icon up and down, raising and lowering the fundamental frequency of the vocal region, relatively suggesting more male or female.

I am fascinated by how I perceive Q's voice. When confronted by ambiguity, we want to remove uncertainty and apply a label, to decide, definitely, what is going on. Of course applying labels is sometimes not the best course of action. Labels, like pigeonholes, sometimes just don't work. I think we can agree that mislabeling something is never good.

Maybe it's another one of the eternal mysteries of sound and how we perceive it. Everything we hear is simply a tickle inside our brains. And based on the sum total of everything we've ever heard, probably from before we are even born, to the present moment, we decide what the tickle means.