I had no idea what to expect. At one point it almost brought tears to my ears. Yet here's the thing: It's (almost) completely fake.
The above was generated automatically from Google's Notebook LLM - from just two pages of this website - my about me page and a song I wrote from an abandoned book. It's an AI voice application that can summarise any text into voice.
So what do I think of this?
It really taps in emotions:
First of all, it's the timbre of the voices. Listen to how the woman broadcasting pauses and returns in almost a whisper. The male voice is deep and reassuring. I've used many voice to text applications over the years, but they never escaped the robotic voice. This sounds human.
Secondly - hearing my life told in such a way, was riveting. The intro I wrote for this website was something I reduced to a digestible nutshell. But upon hearing this, it seemed to fill in the gaps, that I was missing, a human voice telling my story.
Granted the voices are cultural - it's quite American, and there is absolutely nothing wrong with that. I imagine if were British or even my native New Zealand, the approach on the subject might have been a bit more restrained, and maybe even a dash of cynicism? This felt like I was on Oprah, telling my life story.
The podcast made me feel like I was on this show.
Notebook LLM makes you feel like a guest on this show.
I haven't made it as a musician, and this made me feel like I was on top of the world. Like I conquered a mountain. And damn, if that doesn't feel good.
It gets information (mostly) right:
From the two pages on my website - it got remarkably a lot. The AI summary cleverly fills gaps by asking questions of what if and why. This sparks the imagination of the listening.
What it doesn't get right can be forgiven:
It mistakes me for a musician who is much more famous or successful. I haven't even recorded any songs yet - though that is in the works. This site might be public facing, but is essentially an anonymous island in a thick sea of information. That's fine just fine by me, but
It doesn't take into account adjacent information. The podcast mentions that the song 'Stars in the same sky' was written from a book that I was writing. Well, this book was abandoned, but you need to read the post about that.
Its strangely terrifying
I've done podcasts before as a guest, years before. I either sounded like a bumbling fool, or unhinged. But this podcast expresses something deep and satisfying - yet these hosts are artificial.
So many questions could be asked.
If AI manages to capture speaking voices, what more will it do singing voices?
If AI can manipulate emotions so well, can this innocently be used for nefarious purposes?
If the source information is wrong or incomplete - can this mislead the listeners?
I foresee this tool can be used masterfully by those that know how to pull the strings. It's a dawn of new creativity, but a creativity that will eat itself much like a ouroboros.
