> When she moved over to AI, though, it was immediately clear that there was something very wrong. “There were no Black people — literally no Black people,” says Gebru, who was born and raised in Ethiopia.
Is that the problem with AI?
> Fast-forward two years and LLMs are everywhere — they’re writing term papers for college students and recipes for home chefs. A few publishers are using them to replace the words of human journalists. At least one chatbot told a reporter to leave his wife. We’re all worried they’re coming for our jobs.
That sounds like a different, and more universal, problem.
An entire hugely important potentially world changing field not having representation from a huge swath of the population certainly seems problematic. That lack of representation combined with the bias present in the training data is a universal problem. And in terms of the second part; the problems with such widespread usage are many fold, with detrimental bias and lack of representation as one of the many folds.
> An entire hugely important potentially world changing field not having representation from a huge swath of the population certainly seems problematic.
I assume lots of groups are underrepresented, depending on how you slice up humanity. Why is that a problem in and of itself? Ensuring training data is representative is a fundamental technique of data science that you can implement regardless of skin color.
> And in terms of the second part; the problems with such widespread usage are many fold, with detrimental bias and lack of representation as one of the many folds.
But what does that have to do with who is working on the AI systems?
The upshot of all of this is that the people who hired these women are now far behind those who didn't.
These people are bad for your organization if you want to make things for humanity. Best to ensure you never hire them, but if you fire them quickly that's all right too.
The reality is that making the tool was far more beneficial to people (including those underrepresented) than to wait till the tool represented all.
5 comments
[ 0.24 ms ] story [ 22.9 ms ] threadIs that the problem with AI?
> Fast-forward two years and LLMs are everywhere — they’re writing term papers for college students and recipes for home chefs. A few publishers are using them to replace the words of human journalists. At least one chatbot told a reporter to leave his wife. We’re all worried they’re coming for our jobs.
That sounds like a different, and more universal, problem.
I assume lots of groups are underrepresented, depending on how you slice up humanity. Why is that a problem in and of itself? Ensuring training data is representative is a fundamental technique of data science that you can implement regardless of skin color.
> And in terms of the second part; the problems with such widespread usage are many fold, with detrimental bias and lack of representation as one of the many folds.
But what does that have to do with who is working on the AI systems?
These people are bad for your organization if you want to make things for humanity. Best to ensure you never hire them, but if you fire them quickly that's all right too.
The reality is that making the tool was far more beneficial to people (including those underrepresented) than to wait till the tool represented all.