The time it takes for a sound to hang in suspense like that is called decay time. Or. Reverberation time.
It’s half life will depend on the acoustics of course. Some sounds will last longer than others. If the sound or the word is pronounced in a room with perfect acoustics and the listener is waiting to hear in the perfect position the sound can hang in the air like a snow flake. Acoustics.
Sound – production, transmission, and effect. Must not be underestimated.
So too, or maybe more so, this applies to the words we use. The evolution of our languages.
Words – their production, their transmission and their effect.
A word that is chosen, framed, pitched and perfectly placed in the ear of the recipient or even read from a scrap of paper has a reverberation time way past its sound. We use words as though there is an endless supply, as though they are cheap, as though they are replaceable. As though their meanings can be assumed. But they are not cheap. They are quite specific. There are many words in the languages we choose to use and they have impact. Words are designed to have impact. I cannot take back a word I have written that you have read. I cannot return it. We need to be careful.
We must choose wisely from our words, protect them, wrap them in kindness with intent and study their impact because one word can reverberate through a day or a lifetime.
Enter AI.
In AI development there is a particular job called data labeling. A company will be engaged to describe in English (the language of coding) a particular image or motion. We do similar stuff when we label an image right here in WordPress. These sentences, these collections of words are then used to train the Artificial Intelligence creating a program to recognize and use this image or motion on other stages. Do this well because our words and descriptions, are being scraped from the internet to train AI.
But many companies engage other companies to label images and sounds specific to their industry. (Instead of stealing our image descriptions from the internet). The label is the vehicle that moves the data. These are massive undertakings with hundreds of thousands of images (hopefully ethically acquired) to be labeled – in English. But the companies who are contracting out these jobs might not have massive budgets. So they look to the third world and are employing people who struggle with English as a second language. (because they are cheaper) to create the labels. The questions of ethics, and survival and human exploitation aside, this leads to some very dubious image descriptions. For example instead of writing: ‘the hesitant fox jumped over the log’ the sentence might be ‘the furry dog turned over the log, slowly’.
There is little quality control – once again due to budget restrictions. The people who head up these companies definitely have life styles we can only guess at with their first class travel and flash cars. They might not care about the words. Maybe they do not care to spend the money to ensure they are right words for our future.
And once these image descriptions are keyed into a program there are way too many to track and correct. Words begin to change their emphasis. The meaning of the image is skewed. The English or Spanish or French becomes diluted because this machine – this AI program – will roll forwards using words it has been taught. Matched with meanings that are not correct.
The impact is quite terrifying.
In this stage of training AI – it might only be another year or so because the work is at a frantic pace now – as our film industries and health care systems and travel and music and art is slowly fed to AI, we must be careful of the words we use to label our images, the words we frame and carefully place in the ear of the machine. How we train AI will impact our language going forward.
I think. What do you think?
It feels like the ultimate payback that the lowest paid struggling masses who have been historically exploited, are going to be labeling our future. Creating collections of words that will create collections of actions. The peoples who do not speak English at home are impacting language itself.
A new kind of decay.
Celi
The Kitchen's Garden 
19 responses to “Decay Time: The Reverberation of Words”
So far, IMHO, what AI produces is extremely boring …but humans have become somewhat homogenised and boring.
And I rest my case!! Clever.
I’m all about the human factor, something is lost in translation, and it’s not a good loss.
We need to make sure our own contributions use full rich language. I am not good at this actually. Often I don’t take enough care with the words I put out there .
No worries C! We all tend to take a few shortcuts every now and again, but the thought process is irreplaceable (and unmistakable; )
At least right now AI is only able to go as far as the human factor behind it is capable but how long will that last and what will we have to endure in the process- like more false narratives being pumped out. I wonder how long it will be before we begin to distrust everyone we thought we knew and as AI gets better will we still be able to tell the difference.
from my discussions with the film people here it has already gone way past what we thought was possible 6 months ago. Thousands and thousands of animators will be out of a job.
I didn’t want to hit the ‘Like’ button. I needed a ‘$&@/ yes’ button instead.
like I’d a stupid button. We need to write a letter and get new buttons!!
I really like words, some more than others. Some just evoke amazing images and I think you need a real brain for choosing those words. And of course it’s good to remember you never have to eat words not said.
There is so much to think about in this comment!
Remember that AI relies only on written or spoken words, images, and sound. The full range of senses available to human beings, or even animals, far exceeds anything artificial technology can produce. Also, real life is capable of spontaneity, a talent that can’t be pre-programmed.
totally agree. Here’s a question though. If you heard a simply beautiful piece of music. Perfectly rendered and played with heart. And you loved it. Would you stop listening to it after you discovered that it was AI generated with sounds scraped, labeled and regenerated from the internet using another AI program.
But it is good – really good. And you have been listening to this composer for 6 months now. Would you still love it? Or feel tricked.
I don’t know. I’ve felt tricked over and over by humans, too, and by false expectations. I don’t feel strongly enough about anything to be swept away.
I do worry that lazy people and cost-cutting corporations will depend on AI too much and there will be more misinformation out there. It’s like all technology…it can be used for good or for evil. Any way you look at it, though, it will change the world. Thanks for the thought-provoking read, Cecilia.
I do worry that lazy people and cost-cutting corporations will depend on AI too much, leading to more misinformation out there. Any way you look at it, it will change the world. Thanks for the thought-provoking read, Cecilia.
I think if we can keep our content rich, the output will be less homogeneous. English is full of wonderful, nuanced, descriptive words. The trick will be retaining them all if AI becomes the master content generator. Can we feed AI a dictionary or two? I’ve always veered away somewhat from labelling my images because then they cannot be searched and hijacked. It hasn’t stopped other content creators pinching the occasional shot, but might it stop wholesale adoption by “intellects vast and cool and unsympathetic” (HG Wells, War of the Worlds)?
I am thinking that I want an AI search to find my descriptions – at least we know that one will be right!
Certainly not in order of importance, but lots of nouns follow – and many more missing, I’m certain; but: cadence, enunciation, expression, emphasis, pronunciation, sentence structure – all of which we are consciously aware of while listening (or not; )