Ai-powered Content Moderation And Its Effects On Internet Freedom – OpenAI recently shared that it is working on a content manager customization system using GPT-4 LLM. The system will use artificial intelligence (AI) to make policy reviews faster, make consistent decisions and reduce stress on human regulators.

Traditionally, content moderation has been an intellectual and labor-intensive process, with human moderators responsible for sifting through large volumes of content and filtering out harmful material. Currently, this filter design is time consuming and in some cases frustrating for moderators.

Ai-powered Content Moderation And Its Effects On Internet Freedom

Another advantage of using GPT-4 in an AI-based content management system is the ability to interpret and modify complex content policies in real-time. It’s something that historically took months to complete and execute, and it can all be condensed into hours using artificial intelligence.

The Flaws In The Content Moderation System: The Middle East Case Study

That said, OpenAI still recommends human control, at least to begin with. For example, after policies are written, policy experts work with GPT-4 to refine policies until they meet quality standards.

Simply put, the benefits of an AI-powered system controlled by a human operator can increase stability, speed, and protection for the mental health of human operators. This report explores the role of artificial intelligence (AI) in social media content moderation. . These are: (1) why artificial intelligence will play an increasingly important role in regulating online expression, and (2) why even a moderate-minimalist starting point cannot avoid regulation and all the risks and compromises that come with it. Below are (3) some examples of AI’s potential role in the regulatory process and some of the conditions implied in each type of use, (4) why expecting too much of AI or ignoring its fundamental shortcomings can be frustrating, and lead to greater injustice. .

Using AI to Curate Online Content: Or the tech industry invested in automation and what I got was a questionable decision.

What Is Content Moderation [10 Things You Must Know]

In March 2021, in one of the many social media hearings of Congress, Mark Zuckerberg, CEO of Facebook (now Meta), said the following.

More than 95% of the hate speech we’ve removed is done by artificial intelligence (AI), not humans. . . . And I think 98 or 99 percent of the terrorist content that we take down is identified by artificial intelligence, not humans.1

These seem like high percentages. For a long time, tech companies have touted artificial intelligence (AI) as the easiest solution to content curation. (To be fair, some of the moderators were asking, “Are we there?” for a bit.) Then suddenly—faster than I thought, at least—it seems like these magical robots are getting pretty good at fixing the anomalies of human expression. . That’s why . . . problem solved? Have we welcomed this age of Pax Machina and started posting cat memes and gold shillings?

Azure Ai Content Safety

Well, let’s back it up first. Why do we use artificial intelligence to make decisions about online expression? Why should we leave something so important to non-humans? What exactly do we use it for? how good is it How we do it

Is it good? How good is good enough? What if it doesn’t turn out better than we thought?

– infinite scope of online conversation. Scale is the main driver of the web, at least in its current ad-based form, and perhaps in all its incarnations. It’s impossible to absorb the dynamics of running a digital platform without spending time and serious meditation on the wonderful, fascinating conversation we’re having: 500 million tweets a day, 200 billion tweets a year.

Openai Utilizes Gpt 4 To Develop Ai Powered Content Moderation System

I could go on. Expressions that were previously ephemeral or limited in the realm of natural law and the pre-digital printing economy can now spread and circulate globally. Whenever possible, we like to hear ourselves talk.

This is not surprising, as most of us (those who did not own a printing press, newspaper column or press secretary) were, to say the least, waiting for a large part of the story to be spread widely, cheaply and without passing. Compared to previous options, such as writing a letter to the editor or buying some ads, this is a good time to socialize, gain fame, or find a paying job. It’s a win if you like the expression of people who haven’t had access to wide distribution channels (which has brought with it new risks of doxxing, hacking and harassment, and that’s not a win).

With social media and the dangers it poses today, it’s easy to forget what it was like. Despite their evils, the age of digital technology has brought profound changes in communication (of course, based on asymmetric changes in surveillance and coercion). The Internet, at its best and worst, is an age of perpetual curiosity, and the consequences are still surprising for learning more about ourselves and each other (even if none of us knows as much as the companies or governments that track and store the data). . . But at the same time: 500 million tweets a day and 720,000 hours of video? Changes in the nature and dynamics of human relationships will have several serious consequences.

Content Moderators At Youtube, Facebook And Twitter See The Worst Of The Web And Suffer Silently

So we’ve been crafting self-expression benders for over a decade. So what? Why would anyone want to remove it through content moderation, whether done by AI or humans alone? Why don’t we celebrate and let people or the market or God sort it out? Every reader applies his reason to what he reads and sees, and is patient with what he disagrees with, and people grow thick-skinned against evil words, and an invisible voice directs the supply of intelligent thought to a place of suitable demand. There is no god, no moderator.

The question of what to regulate, how much, and the pragmatic and philosophical implications of these options is a giant can of worms that deserves its own book. For our purposes, let’s start by thinking more than a minimum viable regulation, namely that the service provider should only intervene in illegal situations. As we begin to try to implement this limited and illegal playbook, we quickly run into a major question: illegal according to whom?

Courts do not give weight to the behavior and statements of most people, which is a good thing. Trying every damn article online is more expensive and time consuming than any of us can imagine. Some real-world minimalists (eg some Latin American lawmakers, the Indian Supreme Court, the Electronic Frontier Foundation, etc.) say they delete online posts without a court order to protect them from the inconvenience and expense of litigation. as much expression as possible. Now that’s free speech. Today, for practical reasons, this position is quite rare, as will become clear if one tries to implement it.

The Solution To Online Abuse? Ai Needs Human Intelligence

Understanding why minimum viable content regulation is difficult requires careful consideration of two related concepts: human subjectivity and how much is judged outside of court in the world.

Very few types of content or behavior are clearly illegal. For example, material sexual abuse of children is certainly illegal; The hardest part is when you find a false positive you can identify it without ruining people’s lives. But for most types of content, it’s less obvious. Figure 1 shows a useful diagram by Stanford Law School professor Daphne Keller to illustrate the problem.

In most cases, the determination of illegality involves the application of highly subjective factors, using good old-fashioned legal judgment, assumptions about how courts might decide, and risk management. For example, in the United States, there is no statutory protection against secondary liability for trademark infringement. So if someone uses your service to post something, tell them that someone else is accusing them of trademark infringement, such as promoting or selling fake products. Is it illegal to upload that first video? Well, it depends. You should conduct a reasonable investigation of the facts and then determine whether the materials are infringing. In other words, you will do what corporate lawyers do everywhere, which is to assess the legal risks and then do what you can to mitigate the risks for your company based on a cost-benefit analysis.

Algospeak: A New Language To Circumvent Ai Powered Content Moderation

“It depends” is part of what’s wrong with people who think it should be easy to blatantly regulate illegal content. Most decisions about what constitutes “illegal” online conversation are out of court, so it’s a guesstimate, and in most cases there’s no telling what the courts will say. In the case of an online article, the poster can be the person or company that the article mentions or influences.

Reconstruction and its effects, should government control internet and its content, alcohol and its effects, ai on the internet, ai and its applications, tourism and its effects, content moderation and free speech, trauma and its effects, content moderation on social media, internet and its effects, stress and its effects, globalization and its effects