Installed Google Keep, Google’s note-taking application, on my tablet. Mostly because I grew pissed at my inability to choose between the other products at the Android store. (Having, before this, grown pissed of Samsung’s own note taking app, which is sufficient for my uses except for how it sets a maximum on the length of your notes and the total number of notes you can keep. As far as I can see, not because there’s no more space; just because the app’s designed that way. Tablets are computers designed to not be used too hard, grrr.)
Noticed that Keep could take voice-notes. Thought to try this, grabbed a book off my desk, and intoned:
Daniel Dennett the most crooked think of something Boulevard a penis look for East famous for changing an accident on the door please books include gravestones bring children into room go to explain it going stranger 30th and feed them both he lives in North Andover Massachusetts
Er, that’s what my Finnish accent and voice recognition software do. Also possibly a pitch for a horror novel. I meant to say this, off the back flap of Intuition Pumps:
Daniel Dennett is one of the most original and provocative thinkers in the world. A brilliant polemicist and philosopher, he is famous for challenging unexamined orthodoxies. His books include Brainstorms, Brainchildren, Elbow Room, Consciousness Explained, Darwin’s Dangerous Idea, and Freedom Evolves. He lives in North Andover, Massachusetts.
What Google’s voice software is doing here is very clever. I don’t think I can pronounce Massachusetts the right way; but it’s probably so alone in the landscape of existing words that it’s easy to pick up. North Andover? That’s not lonely; but Google no doubt has an algorithm for scanning “blurble blurble Massachusetts” for Massachusetts places that sound something like “blurble blurble”. (And since North Andover was capitalized, it was specifically recognized as a place name!) The same for Daniel Dennett; Google’s not depending just on what I say, but also on what it knows about words.
The next step would be Google assuming you might be saying something that’s already written, and searching the whole Internet for matches; then it could find a copy of that bio and recognize it near-perfectly, punctuation and all. (Perfectly except for where I dropped a word or said one so badly Google thought I was remixing things.)
Well, the next step unless this would lead to Google overassuming, and cribbing some online BDSM fiction instead of what I was saying.
Even if I wasn’t reading off a book flap, there might be sentences that are often used (well, none in that example) and thus easy to recognize; Google’s sitting on the biggest cache of language usage in the world. (Wait, some of the sentences in my sample are easy to recognize: the list of book names isn’t a likely stream of sounds, yet it’s very likely to match at multiple points a list that’s floating in many places online, close to recognized phrases “Daniel Dennett” and “North Andover Massachusetts”. Then you look at the uncertain matches, see they could be the other book names — and bam!)
Before this try, tried something freeform on testing the thing and not feeling optimistic. Google Keep recognized the words “Google” and “fucking” and bleeped one of them. Oh well.
Now, some more names.
baby you didn’t check the script for pictures from Harry’s be to Myers much you
Er, no. I meant —
Denny’s didn’t Richard Lucas Christopher Hitchens send Harry’s Myers my room temperature
Denny’s didn’t Richard Dawkins Christopher Hitchens sam Harris be checked Myers Malayalam not much you
Close enough. Tried to say this random but directed list of names:
Daniel Dennett Richard Dawkins Christopher Hitchens Sam Harris PZ Myers Maryam Namazie
Clearly me and Google still have ways to go.