this post was submitted on 15 Aug 2023
83 points (97.7% liked)
World News
32285 readers
1112 users here now
News from around the world!
Rules:
-
Please only post links to actual news sources, no tabloid sites, etc
-
No NSFW content
-
No hate speech, bigotry, propaganda, etc
founded 5 years ago
MODERATORS
you are viewing a single comment's thread
view the rest of the comments
view the rest of the comments
I completely disagree. It absolutely is AI doing this. The point the article is trying to make is that the data used to train the AI is full of exclusionary hiring practices. AI learns this and carries it forward.
Using your metaphor, it would be like training AI on hundreds of excel spreadsheets that were sorted by race. The AI learns this and starts doing it too.
This touches on one of the huge ethical questions with regulating AI. If you are discriminated against in a job hunt by an AI, who’s fault is that? The AI is just doing what it’s taught. The company is just doing what the AI said. The AI developers are just giving it previous hiring data. If the previous hiring data is racist or sexist or whatever you can’t retroactively correct that. This is exactly why we need to regulate AI not just its deployment.
"This touches on one of the huge ethical questions with regulating AI. If you are discriminated against in a job hunt by an AI, who’s fault is that?" It is the fault of the company hiring practices, which are to blindly trust an AI without testing whether or not it is discriminatory. It is also the fault of the producer of that AI software (or service) sold to the company for screening candidates. No new laws are needed to hold either of them accountable, existing laws cover the ground well. That company selling the AI screening services could have just been called "crystal ball hiring" before AI and would be equally liable if they just discriminated in their hiring suggestions. The tool isn't the thing that needs regulating, the actions people and companies take based on the tool is. And that is already well regulated.
Make an AI in the privacy of your own home that does ____ literally anything? Fine. Collaborate making an OSS AI to do whatever with some of your friends? Also fine. Sell that AI as a employment screener app? Better make sure you've tested it to not have discriminatory outcomes. Use that AI to screen employees? Same deal.
The “AI” is simply a program used to filter data. Whose fault is it if using it causes problems? The nitwits who choose to use these programs and trust the results without understanding the limitations.
The "guns don't kill people, people kill people" argument, eh?
Not really the most precise analogy, but sure. Individuals and companies are responsible for actions, not inanimate objects. However, that line of thinking is used as an argument against gun regulations. “AI” is a much broader field and more general tool than a firearm, and also newer and less well understood. People seem to be on average confused by the hype and unclear on what these systems can do well, or not, and what are appropriate uses. I’d hate to see legislators who know barely anything about technology write laws to restrict it prematurely, or more likely, adopt laws written directly by companies and industry groups who could have various hidden motives. For example we have seen both Sam Altman and Musk mention restricting AI research, and in both cases, the real motive seemed to be not safety but obtaining an advantage for their own companies.
For now, I agree with the other poster who said existing laws are sufficient. If a company was to discriminate in hiring, it doesn’t matter whether they used a special program to make the decisions or not.