dylan_brown30•10d agoProlific Poster

Finally got my small research group to agree on AI transparency rules after months of debating

It took us 3 full meetings at the local library in Austin, but we finally hammered out a basic framework for how we document training data sources. One guy was dead set against sharing any code, but after we showed him that MIT study about bias in healthcare algorithms from last year, he came around. Now we're planning to test it on a small project tracking local real estate listings. Has anyone else tried getting a volunteer group to follow ethical guidelines and actually make it work?

3 comments

3 Comments

shane_hayes10d agoMost Upvoted

Oh man, I gotta respectfully push back on this. I think full transparency can actually hurt more than it helps, especially for small volunteer groups like yours. Look at what happened with that open source facial recognition project last year - they published their whole dataset and suddenly it was being used by police departments in ways they never intended. For your real estate listing project, do you really want competitors or scammers seeing exactly how you're scraping and filtering data? Sometimes a little opacity protects the work, like how chefs don't hand out their full recipes. Plus, all that documentation takes time away from actually building something cool.

blake7928d ago

Totally agree with you there, the facial recognition example is a perfect warning for why blind transparency can backfire badly. A little healthy secrecy keeps the good work from getting twisted into something nobody signed up for.

jessej239d ago

So you're saying small groups should hide their methods to stay safe?