pull down to refresh

Maybe you've heard the myth that the QWERTY keyboard layout was invented to slow typists down. The story goes that typists were too quick for mechanical typewriters and so someone came up with the idea of spreading out commonly used letters to slow things down. Today I learned that this is false.
In reality, the QWERTY layout was implemented to keep parts of the typewriter from jamming, but it also sped typists up because alternating hands while typing is faster than words that must be typed all with one hand.
Anyhow, this article talks about how keyboards (QWERTY or otherwise) are the bottleneck between us and LLMs, and how it would be far faster convey context and instructions to LLMs via our voices than a keyboard.
He makes a pretty good case that
Users will gravitate towards the AI system that allows them to input the most context with the least friction, making voice a superior input method for complex tasks.
And he even comes around to the reason he is probably wrong about this thesis: it's embarrassing to talk to your computer in public. Even if it isn't embarrassing, it is not always the case that you want other people to hear what you are working on. And since we are often working in close proximity to others, typing is probably going to remain the dominant form of communication with our devices and the software that runs on them.

I still want to write a scifi story about a world where voice communication with computers (or maybe just AI agents) has become commonplace and there are some people who are particularly gifted at hacking via voice commands. I imagine it as some distorted version of natural language and code combined in verbal form. It would be a little bit like such people were speaking spells which would make these hackers something more like wizards or warlocks...
220 sats \ 3 replies \ @k00b 29 Aug
Anyhow, this article talks about how keyboards (QWERTY or otherwise) are the bottleneck between us and LLMs, and how it would be far faster convey context and instructions to LLMs via our voices than a keyboard.
A friend of mine was fond of saying the real bottleneck on human communication is the speed of thought. He'd say that whenever we'd discuss more direct brain-computer interfaces. In that vein, I wonder if producing more tokens with speech would actually yield more information for LLMs or just more noise.
reply
If we ever all get brain connected I'll be taking all of the drugs at the same time and make y'all deal with that.
reply
68 sats \ 1 reply \ @kepford 29 Aug
I have no real knowledge on the speech to text world but based on my use of it over the years it has a long way to go to beat typing.
reply
Agreed
reply
100 sats \ 0 replies \ @k00b 29 Aug
wizards or warlocks...
reply
0 sats \ 0 replies \ @OT 22h
I'll take a guess that the next device to supercede the smart phone will determine whether we continue to type or not. Imagine a smaller Rabbit that you could clip onto your shirt. Interactive glasses could be a possibility too.
t's embarrassing to talk to your computer in public
Being able to switch between talk mode and thought mode would solve this.
reply
0 sats \ 0 replies \ @freetx 23h
Not sure about that....whenever I engage "voice mode" on any of my LLM interfaces, I don't seem to get as good results long term.
I think thats because thoughts leaving your mouth are not as structured as when you finish typing them.
Typing gives you the chance to edit, which is crucial for maximizing intent.
Slop in -> Slop out
reply