Why are you even using genAI?
November 14, 2024•743 words
A friend pointed me in the direction of this article as yet another example of AI bias in action:
https://community.thefemalelead.com/p/all-i-wanted-was-an-ai-image-of-a
It is really important that awareness is raised of the weaknesses of genAI and especially the way it will always perpetuate bias and structural injustice in society. After all, it is just a ‘next token generator’ and if the previous tokens had a pattern, however unfortunate that was, it will continue the pattern.
But I also had a different reaction. I couldn't understand why the author had tried to get the image she needed with an AI image generator in the first place. I put 'Man cleaning a toilet' into a good old-fashioned image search and hundreds came up which met her requirements. Sure, most were stock images that needed a small fee paying - and I will come back to 'cost' in a minute - but even restricting the search to 'royalty-free' gave loads of usable images.
The author, Dr Pillay, just says 'So I turned to Midjourney'. She is a 'surgeon and a writer' so presumably not a 17 year old who never wrote blogs before AI infected the internet. She knows there are alternatives. One can only guess at the reasons, but I suspect 'quick', 'convenient' and 'free' will be in there. It is hard not to hear these as close synonyms for 'lazy and cheap' - they certainly aren't antonyms.
So let's focus on the 'free'. It isn't free. You pay for using genAI by being a data point which demonstrates the value of the product and thus enables to company producing it to raise more money and eventually be sold at some staggering profit to the original investors. Every time you use genAI you make a billionaire a tiny, tiny bit richer. Remember Google making its money 'one cent at a time'?. The same is true here.
Even worse, if you create 10 or 20 unusable images of a man cleaning a toilet, that benefits them more than you creating just one image which is perfect for their job. And then if you write an article about using genAI and forget to mention that a traditional image search would have been much better, you are adding to the #AIHype which normalises this. And thus makes then even richer as other people use genAI rather than other perfectly good tools.
Contrast this to using an image created by a human and either paying or acknowledging hem. Who benefits from this? Ordinary people who work for a living. Not billionaires who get rich by making unearned windfall profits out of speculative investments and market distortions.1 Oh, and by externalising the true environmental cost of using their tools.
Towards the end of the article, Dr Pillay does give a positive example of her use of genAI:
Despite its flaws, AI has been a powerful tool in my creative projects. I recently made a short film titled ‘An Incontinence Truth’, using AI-generated images, voices, and music to raise awareness about urinary incontinence, without using any real patients’ words or images. Without AI, this film would not have been possible.
I am not sure I believe that claim. It would have been possible. After all, people made films about things like this without using any real patient's words or images before genAI. They paid real human script writers, composers and actors.2 That is 'more expensive' and slower. But maybe those are costs we should be paying rather than taking shortcuts which increase inequality and destroy the environment.
Let me end by being clear this is not a personal attack. Dr Pillay writes important socially aware and beneficial things. As a surgeon and a writer she is a force for good in the world. Much more so than me. But as an advocate of genAI, even a slightly critical one, she is representing a social problem we need to fix, and fast.
-
No one has earned a billion dollars in the normal sense of being paid for their work. If you worked 40 hours a week at $100 an hour, it would take you 5 years to earn a million but 5000 years to earn a billion. You would have to start before they had built the pyramids. ↩
-
People whose work was stolen to train the genAI, so not only are they being put out of future employment, but not even being paid properly for what they did in the past. ↩