I’d like to see the source that says Dragon’s WER in the 90s was 3-5%. I used Dragon in the 2000s and it just wasn’t comparable to the current state of the art.

https://dragon-medical-transcription.com/history_speech_recognition.html, for example. a lot of adverts and awards were given to it (admittedly awards like PC Mag that were probably paid advertising… but that’s why I went with Open AI’s assessment on whisper at 2%.) Dragon was boasting 99% accuracy after (admittedly months) of training; and it frequently reached it. there were some gotchas in that- the months-long training was a big one. The other was that you frequently had to slow down and be careful to enunciate that you don’t have to do with modern systems (including the MS versions of Dragon- they bought it out at some point)

whisper.cpp is an opensource implementation, although I’m not certain exactly how open.

It’s on the MIT license, if that helps. I take issue with anything OpenAI is involved in. for oh-so-many reasons.

You’re only partially correct about input speed. If you want to dictate an email then yes you need to think about each word you want to say and the order in which to say them. Coupled with an LLM that problem is diminished because you can just kind of have a conversation with the LLM and tell it to draft an email.

and how much of that conversation with an LLM is “No, what I want is…” because it assumed something; or just straight up hallucinated or the typo made it go off on a tangent?

As for whisper, I can find sources that are saying for American-English speakers in a not-noisy environment (aka the best case scenario,) the model has a word error rate between 2-8%. For reference, Dragon NaturallySpeaking had a WER of 3-5%. So I wouldn’t say that Whisper has made any substantial improvements, and they’re OpenAi. you can trust them if you want. I don’t think that’ll work out well in the long run, though.

I remember Dragon SpeechDragon Naturally Speaking saying the same thing in the 90’s. It’s improved, but not enough to make it useful as more than an aide for people who can’t type. I do agree, that for simple accessibility, it should be integrated into every field, but I doubt it’s ever going to take over.

As others have noted, that it’s only technically true that dictation is faster than typing. In a practical sense, there’s a fair number of reasons why that’s not the case, including that usually thinking about the entry is what’s the slowest, and also the errors in both are typically what slows people down.

there’s also the problem of, for example, keeping entries confidential. You don’t want to speak your passwords where others can hear you.

Well the mask is a steno mask

Theoretically most people will speak faster than they type. You have to type around 180-200 wpm to be faster than speaking.

(I say theoretically, because usually typing speed ratings also ding you for errors, and uh, speech transcription isn’t really there, either.)

I don’t normally suggest punching people because they wear glasses…. But I think an exception should be made.

They’re very punch-able glasses.

“I’m gonna need a password.”

there’s a huge in the city where I work. Hundreds of birds. Sometimes when it’s extremely cold… they roost in a few trees at my building. They seem to move around a bit; in any case, I knew a guy that made the dumb decision to shout at them.

They did not let him forget his mistake. And, yeah. I’m not sure changing town was enough… dude was a total dick so I’m there for the pettiness.

Like most people, they’re highly motivated by food.

Find where they roost and feed them. Roasted peanuts, etc, are a good option (just avoid super salty stuff.)

Be patient and set the food down and walk away to start, then walk less far then set it down and stay close and so on.

Also remember that they have excellent memory and can recognize individuals, so they will remember you. They might even come find you.

(And don’t be mean. They will remember you! And they might even come find you.)

A rude idiot!

Meanwhile, the end-users waiting for it to come back up:

BlackBerry wore it better!

I hope you oblidged!

Okay. Lots of reasons.

Another is it’s funny and fun and memorable.

It helps in this situation, where the instructions are unclear and thinking through the process in a logical manner.

Fortunately, kids aren’t quite clever in that way.

There’s a reason school teachers are now doing that exercise where they have kids list instructions on how to make a pb&j, and then following them exactly.

It takes a decade or more to build a new chip factory and bring it to full production.

Shorting the supply now means driving prices up and holding, making bank.

(I don’t know if that’s what they’re doing, but there is motive there.)

I’m getting some adorable napoleonic vibes. 1 part big attitude, 1 part short cat. 5 parts adorable.

The last version of windows I willingly used was ME.

In my defense I was going through my rebellious teenage phase. (My dad raised me on red hat.)

Yeah, I can see why that would be. Nebula looks like a riot.