I fell in love with an AI voice
· 4 min · #miscellany
Okay, no, I didn't actually fall in love, but I did feel something weird towards a computer!
I accidentally found out Microsoft's crappy browser, Edge, has a built-in page reader with an incredible, eerie, unsettling, sweet voice. I've let her read my blog posts, and I've been sitting here for half an hour, listening to her reading my own words back to me... what a narcissistic, perverted thing to do, I know.
The thing is, I don't think I've ever heard a computer sound so strangely human and... sentient?, capable of actually feeling what it's saying, since Scarlett Johansson in her.
A little context: I downloaded a Windows 10 ISO and installed it in a virtual machine to try editing a file in Microsoft Word; I can't get myself to use LibreOffice, and Google Docs lacks a few nice-to-have features I needed for this document[1].
Out of curiosity, I opened Edge and navigated to my website to check out how it renders with Windows' default fonts (it's not bad). From there, I accidentally pressed the button that activates the page reader and... good heavens! It's... alive!
I looked it up online, they call this feature Read Aloud.
I swear this is not an ad; I hate Windows and I would never use Edge as my default browser but... that voice! Last time I tried using a page reader was on Firefox not that long ago and that voice sucks; no, wait, it doesn't suck. What I mean is that it sounds robotic, just as you would expect from an artificially generated string of consecutive sounds, right? It sounds like — not to be rude — a bad imitation of Stephen Hawking.
Edge has a ton of voices to choose from, varying for language, gender and age, and most of them sound just a little less fake than that of Firefox. But not Ava. Ava[2] feels like a real person, trapped inside my computer, and forced to read whatever text I give her, and I almost feel the need to set her free.
It's a marvel of engineering, an incredible achievement, born out of a correct use of artificial intelligence, and I'm sure it will be of great help to those who have vision impairment and such issues. Yet, it is also unnatural. It sounds too real, exaggerating the perfection of some pitch movements.
I don't trust her. Her tone is that of one who desperately wants to be trusted, and that is precisely why I don't. Yet, she hypnotized me.
I recorded it reading a few sentences. Take a listen and then tell me if you too felt a shiver.
The way she doubles the "L" at the beginning of "love"[3]; the upward inflection before reaching the question mark; the "No" sounding exactly like she had just realized she'd used the wrong word, and was already trying to find a better way to explain herself; the way she almost whispers "this" because of the following ellipsis. Come on!
I'm speechless (pun secretly intended).
Yes, it still has a few imperfections but just think that it can only get better than this!
I don't think I'm ready for what's to come in the next few years, in terms of AI achievements, if I can't even contain my excitement — and fear — for a fake voice.
To end with the same honesty with which I started writing, I admit this: some time has passed since the first listen and now... it doesn't sound that special anymore; I might have made a big deal out of nothing, and thus wasted both our times.
I realize that, perhaps, the main reason why I felt hypnotized and compelled to keep listening was because she was reading my blog. I had never heard those words spoken out loud before, never with an intonation different than mine. She gave me the opportunity to see what it's like to perceive these pages from a different perspective. It wasn't just about the voice, it was about the words themselves.
Again, probably too narcissistic.