Covert Racism in AI: How Language Models Are Reinforcing Outdated Stereotypes
Covert Racism in AI: How Language Models Are Reinforcing Outdated Stereotypes
Despite advancements in AI, new research reveals that large language models continue to perpetuate harmful racial biases, particularly against speakers of African American English.
You mean the problems that experts said 10+ years ago would happen are happening?
"Wow Johnson, no matter how much biased data we feed this thing it just keeps repeating biases from human society."
Sample input from a systematically racist society (the entire world), get systematically racist output.
No shit. Fix society or "tune" your model, whatever that entails..
Obviously only one of these is feasible from a developer perspective.
While it may be obvious to you, most people don't have the data literacy to understand this, let alone use this information to decide where it can/should be implemented and how to counteract the baked in bias. Unfortunately, as is mentioned in the article, people believe the problem is going away when it is not.
The real problem are implicit biases. Like the kind of discrimination that a reasonable user of a system can't even see. How are you supposed to know, that applicants from "bad" neighborhoods are rejected at a higher rate, if the system is presented to you as objective? And since AI models don't really explain how they got to a solution, you can't even audit them.
True, and it upsets me because we can't even get a baseline agreement from the masses to correct systemic inequality.
..yet, simultaneously we're investing academic effort into correcting symptoms spawned by the problem (that many believe doesn't exist).
To put this another way. Imagine you're a car mechanic, someone brings you a 1980s vehicle, you diagnose that it is low on oil, and in response the customer says, "Oil isn't real." That's an impasse, conversation not found, user too dumb to continue.
I suppose to wrap up my whole message in one closing statement : people who deny systematic inequality are braindead and for whatever reason, they were on my mind while reading this article.
I'll be curious what they find out about removing these biases, how do we even define a racist-less model? We have nothing to compare it to.. another tangent, nope, I'm done. Zz.
I like that they say "outdated" stereotypes like they used to be true but now they aren't.
Come on people, keep your steroetypes current.
Shit in, shit out. That’s AI. You can’t guarantee a single thing it says is true, and you have to play whack-a-mole forever to get it to behave. Imagine knowing this and still investing time and money in it. We could be investing that in education and making the human experience better, but instead we’re stuck watching capitalists harness it to replace people, and shove half-baked ideas out the door as finished products.
Look, I love tech. I’ve worked in tech for 20 years. I’ve built apps that use AI. It’s the one tech that I despise watching capitalists have control of. It’s just chatbots all the way down that don’t know what they’re regurgitating, and eventually they’re going to be vacuuming up nothing but other AI content. That is going to be the future. Just bots talking to other bots. Everything completely devoid of humanity.
Shit in, shit out. That’s AI.
"On two occasions I have been asked, – "Pray, Mr. Babbage, if you put into the machine wrong figures, will the right answers come out?" ... I am not able rightly to apprehend the kind of confusion of ideas that could provoke such a question."
—Charles Babbage, on his analytical engine, 1864
I need to add that to my quotes book, it's great
Replace "AI" with "humans" and this rant is still perfectly coherent.