Imagine a world without keyboards. A world where keyboards are used rarely only by specialists, like a court reporters stenograph or a draftsman’s compass.
While keyboards are found in every home, workplace, and now school backpack, I think this is going to look antiquated in about a decade.
What will replace them? Your own voice.
Sure, you’ve used dictation software before, like Nuance, your phone’s voice recorder, and it all just kind of sucks. Lots of umms, lots of errors and random punctuation thrown in while you pause and think.
But just try an experiment. Take out your phone right now. Find your recording app (Voice Memo on iPhone, Recorder on Android), hit Record, and just describe something you’ve been working on. Some life problem you’ve been trying to solve. Like, who’s driving the boy to baseball tonight, since both you and your spouse are busy. Talk a lot. Repeat yourself. But record it.
Now, take the transcript (both apps automatically generate this), copy it and paste it into your most or least favorite LLM. I usually use Google Gemini, but ChatGPT, Claude, or Grok are fine too. Paste your transcript, hit Enter, and watch what happens.
If you’re like me, that first experience ruins the keyboard.
This is the wave of the future, and good riddance. Your voice is one of the most miraculous parts of you.
“I hate my voice!”
I’m biased. I think humans are separate and above the animals as a creature, and special in the universe.
My bias started growing long ago, but really solidified during my decade with Lyndon LaRouche. One of the things we did that in hindsight looks pretty culty, was sing. We would strive to do a voice warm up every morning as a group, and would practice singing Bach and other classical songs several times a week. Then, we would get together next to one of our tables that howled “9-11 Was An Inside Job!!” and sing beautifully to the passersby.
There were many reasons to sing. For example, it got us used to working together to make something beautiful. It also helped strengthen our vocal cords to talk to a hundred people a day, sometimes over loud traffic and other background noise.
But the main reason we sang is because, as humans, we use our voice to communicate ideas. Singing was a way to teach each other the mechanics of how our voices do that.
Ideas are objects that animals just can’t understand. Maybe they can sense there’s something there, but not as anything more than “there’s something wrong” or “boss is happy”. But you can’t explain something like justice to a dog.
We use our voice to communicate those ideas. Chimps use their voices too, and they do it to express themselves and their needs. But ideas are not sensible objects. They are only sensed in your mind.
I don’t know if our physical vocal instrumentation was specially designed for communicating human ideas, but it fits the purpose really well. A lot of our ums and ahhs are there because we don’t know what to say next. But some are there because they’re supposed to be. It’s those pauses and inflections that don’t have to be there, that we use to “get the idea across”.
When you speak (or sing), you’re creating sensory stimuli intentionally, to create patterns of sensory stimuli on the sense organs of the listener. That listener can pick up on the sounds or silences that aren’t supposed to be there, and use those clues to regenerate YOUR idea in THEIR minds.
When done well, we get forms of art like poems and songs. Percy Bysshe Shelley wrote a pamphlet a couple hundred years ago called “In Defense of Poetry“, where he claimed that the earliest form of human language was poetry. We were testing this vocal tool’s ability to describe things that had never been described before. So, concepts that had no words yet, could still be described using this special ability of the human voice.
AI thirsts for your voice
Did you run that little experiment above, and drop the recording transcript into your LLM de jure?
If so, you’ll find the response is vastly more rich than what you get when you type a prompt. I can bullshit about why this happens, like how attuned the LLM is to tiny details like punctuation, etc etc. But I think the main thing is, when you speak, you tend to follow the natural contours of thought. You’re more likely to say some detail that wouldn’t make it into your typed prompt. You might even just go down a tangent you wouldn’t have expended the finger effort to explore.
All that extra context you provide gives the LLM so much more information to work with. Neither I nor anybody else really knows what’s going on in these models when you give them a prompt. We can wave our hands about “just spits out the statistically most likely next phrase”, and other derivations from the true logical design of these systems.
However, even if that’s the case, speaking ideas out loud naturally and necessarily incorporate concepts that just don’t typically go together, and that a logical system certainly wouldn’t “think” to string together. This makes the output of the LLM become something like a novel, rigorous extension of your paradoxical idea concept to limits you haven’t taken the time to explore. And it does this in seconds.
And, sometimes, it’s just quicker to say “Hey Google, put garbage bags on my grocery list in Keep” than to type it. That’s also a big win.
People were not designed to type words on a device. We were meant to speak them into existence, and these new models make a return to that existence possible.
End of the Keyboard
The keyboard was a compromise – a mechanical bridge that allowed us to communicate with dumb machines. But those machines can now speak Human, and understand the intent of what we’re saying.
As we move toward a world without keyboards, we move into a world where we can reclaim the oral tradition that Shelley and the other poets championed. We are speaking our ideas into the world, one messy miraculous word at a time.
So open your recording software, take your hands off the keyboard, and speak a comment below. The AI will clean up the mistakes.
Leave a Reply