Ranter
Join devRant
Do all the things like
++ or -- rants, post your own rants, comment on others' rants and build your customized dev avatar
Sign Up
Pipeless API

From the creators of devRant, Pipeless lets you power real-time personalized recommendations and activity feeds using a simple API
Learn More
Comments
-
JsonBoa30162y@kobenz 's argument is already solid enough, but here is another line of thinking: although the original training dataset is relevant to the performance of an AI, it pales in comparison with the usage and feedback data from real world interactions, used to train and reinforce newer iterations of the AI.
Thus, to be effective, a training and validation data disclosure policy would require the disclosure of real life user interactions as well, and that is a whole other Pandora box. -
JsonBoa30162y@tosensei , as said @kobenz . And the point is not to make it impossible to use technology for evil, but to make it illegal to make large profits from it.
For example, regulating deepfakes prevents some shithead from copying Hollywood stars to feature on their ads, and getting away with it. It also vastly limits deepfake revenge pornography, since platforms cannot simply ignore the problem.
Regulation and it's proper application using legal means, by itself indeed does not solves the problem of misusing technology, but it is the most fundamental step.
Think hard drugs. Laws and law enforcement did not make those go away. But it could be way worse. For example, if "McMeth" was a product that was allowed to exist and be advertised in billboards. -
JsonBoa30162y@tosensei exactly my point: proper regulation leads to benefits, even if it does not obliterates the problem.
That's why AI and technology in general must be regulated. Otherwise it'll become (even more) organized crime and shoddy coding all over, especially after a prolonged period of no regulation leads to an overcorrective prohibition crackdown.
Would it make sense in order to regulate AI to force companies to publish the dataset they used to train a specific model?
question