Congress takes Aim at Algorithms


“I agree that there must be a responsibility, but I do not think we have found the right words to describe the processes in which we are involved,” said Jonathan Stray, a visiting expert at the Berkeley Center for Human. – AI-compatible that learns dynamic algorithms. “What is nurturing, what is nurturing, what are your preferences, what is a mind?”

New Jersey Democrat Frank Pallone’s Justice Against Malicious Algorithms Act, for example, removes security when the platform “knew or should have known” that it was creating “personal information” for the user. But what counts as human? According to the bill, it uses “personal information” to support the popularity of some things over others. That’s not a bad meaning. But, on the face of it, it seems to suggest that any platform that does not show anyone the same will lose the security of Section 230. Even showing someone the records of the people who follow them no doubt depends on the person’s knowledge.

Malinowski’s bill, the Protecting Americans from Dangerous Algorithms Act, may invalidate Article 230 for security reasons by terrorists if the platform “uses a method, model, or other calculation method to classify, summon, promote, approve, enhance, or modify information or representations. of information.” It contains exceptions, however, for algorithms that are “predictable, clear, and visible to the user,” and list other examples that may be relevant to the bill, including feed-based feeds and rankings based on popularity or user feedback.

There is a great idea for that. One problem with engagement-based algorithms is their invisibility: Users are unaware of how their personal data is being used to target them and what the platform predicts to be connected to. But Stray also said that distinguishing between good and bad algorithms is not easy. Selection by user feedback or up / down-voting, for example, is a challenge in itself. You do not want a post with just one vote or a five-star comment to get to the top of the list. The standard way to fix that, Stray explained, is to calculate the error of the given item and place it according to the bottom of the distribution. Is that approach – which took Stray a few minutes to explain to me – clear and concise? What about the most important thing as a spam filter?

“It is not clear to me whether the goal of eliminating systems that are ‘simple’ enough would completely eliminate any system that is effective,” Stray said. “My skepticism is, probably not.”

In other words, the interest rate that removed Section 230 security in relation to algorithmic concepts may be viewed as a direct refund, especially in terms of social networking sites. Jeff Kosseff, author of Section 230 certified book, Twenty-six words made up the internet, added that internet companies have a lot of legal protection to return, including the First Amendment, even without legal protection. If the law is full of differences, and with the exception, the companies may decide that there are simpler ways to defend themselves in court.



Source link

Leave a Reply

Your email address will not be published. Required fields are marked *