Older people such as myself tend to hate voice-to-text I think because it was so awful in the past. And if you screwed up with it in the past it was a less understandable excuse that “I was using voice-to-text.” And because we were all forced in some way to learn to type well.
Voice to text works a little bit better now. And I think younger people know everyone else uses it and to forgive when it screws up.
I’m going to need significant levels of convincing. Computers have always preferred specificity and accuracy, it’s half the reason I’m in my current position (MSP Escalations/level 3, half of my success at fixing issues is being extremely specific in looking up exact error messages instead of paraphrasing).
This isn’t a defense of AI; on the contrary, it’s my doubt that AI can read intentions/inflection/emotion better than just writing out what you actually want.
Deepseek recently published a paper in which they describe that vision tokens contain more information than text tokens and that this can be used to compress context.
We present DeepSeek-OCR as an initial investigation into the feasibility of compressing long
contexts via optical 2D mapping.
Experiments
show that when the number of text tokens is within 10 times that of vision tokens (i.e., a
compression ratio < 10×), the model can achieve decoding (OCR) precision of 97%. Even at a
compression ratio of 20×, the OCR accuracy still remains at about 60%. This shows considerable
promise for research areas such as historical long-context compression and memory forgetting
mechanisms in LLMs.
It reminds me of LLM caveman speak, it used to have another option to use Chinese instead of English. A language like Chinese is seemingly better at encoding information in fewer tokens and I think this is the same mechanism why OCR tokens work so well.
That said, I also doubt that voice messages are more efficient than text prompts, but it’s best not to waste too much time engaging with these sorts of LinkedIn posts (and LinkedIn in general).
Isn’t it cheaper and quieter to just type out your prompts?
This is akin to people who have conversations on speakerphone in public places.
Older people such as myself tend to hate voice-to-text I think because it was so awful in the past. And if you screwed up with it in the past it was a less understandable excuse that “I was using voice-to-text.” And because we were all forced in some way to learn to type well.
Voice to text works a little bit better now. And I think younger people know everyone else uses it and to forgive when it screws up.
I hate it because it’s generally intensely US-centric. Not understanding non-US accents or terms.
you assume that vibe-coders can actually touch type. or type at all.
Or be capable of critical thinking.
I assumed they were all working from their iphones
AI needs to measure the level of confidence in your voice, to calibrate its bullshit accordingly
Speaking is faster than typing, I guess?
Maybe for Eminem.
apparently audio and images are more efficient compared to text for multimodal models?
I’m going to need significant levels of convincing. Computers have always preferred specificity and accuracy, it’s half the reason I’m in my current position (MSP Escalations/level 3, half of my success at fixing issues is being extremely specific in looking up exact error messages instead of paraphrasing).
This isn’t a defense of AI; on the contrary, it’s my doubt that AI can read intentions/inflection/emotion better than just writing out what you actually want.
Deepseek recently published a paper in which they describe that vision tokens contain more information than text tokens and that this can be used to compress context.
It reminds me of LLM caveman speak, it used to have another option to use Chinese instead of English. A language like Chinese is seemingly better at encoding information in fewer tokens and I think this is the same mechanism why OCR tokens work so well.
That said, I also doubt that voice messages are more efficient than text prompts, but it’s best not to waste too much time engaging with these sorts of LinkedIn posts (and LinkedIn in general).
LLMs don’t need accuracy. This just boils down to speaking being faster than typing, especially if your thought isn’t fully formulated.
deleted by creator
As far as I know, these workflows typically involve a transcription model to convert the audio to text, and then passing the text to the model.