Most Americans agree that misinformation, hate speech and harassment are harmful and inadequately curbed on social media through current moderation practices. In this paper, we aim to understand the discursive strategies employed by people in response to harmful speech in news comments. We conducted a content analysis of more than 6500 comment replies to trending news videos on YouTube and Twitter and identified seven distinct discursive objection strategies (Study 1). We examined the frequency of each strategy's occurrence from the 6500 comment replies, as well as from a second sample of 2004 replies (Study 2). Together, these studies show that people deploy a diversity of discursive strategies when objecting to speech, and reputational attacks are the most common. The resulting classification scheme accounts for different theoretical approaches for expressing objections and offers a comprehensive perspective on grassroots efforts aimed at stopping offensive or problematic speech on campus.

There is growing interest in ensuring that large language models (LLMs) align with human values. However, the alignment of such models is vulnerable to adversarial jailbreaks, which coax LLMs into overriding their safety guardrails. The identification of these vulnerabilities is therefore instrumental in understanding inherent weaknesses and preventing future misuse. To this end, we propose Prompt Automatic Iterative Refinement (PAIR), an algorithm that generates semantic jailbreaks with only black-box access to an LLM. PAIR -- which is inspired by social engineering attacks -- uses an attacker LLM to automatically generate jailbreaks for a separate targeted LLM without human intervention. In this way, the attacker LLM iteratively queries the target LLM to update and refine a candidate jailbreak. Empirically, PAIR often requires fewer than twenty queries to produce a jailbreak, which is orders of magnitude more efficient than existing algorithms. PAIR also achieves competitive jailbreaking success rates and transferability on open and closed-source LLMs, including GPT-3.5/4, Vicuna, and PaLM-2.

A new open source AI image generator capable of producing realistic pictures from any text prompt has seen stunningly swift uptake in its first week. Stability AI's Stable Diffusion, high fidelity but capable of being run on off-the-shelf consumer hardware, is now in use by art generator services like Artbreeder, Pixelz.ai and more. But the model's unfiltered nature means not all the use has been completely above board. For the most part, the use cases have been above board. For example, NovelAI has been experimenting with Stable Diffusion to produce art that can accompany the AI-generated stories created by users on its platform.

While AI can perform incredibly well with tasks that have clear parameters, such as a game of chess, humans are still better at making the tough calls and dealing with unpredictable situations. Gray shares the example of Uber wanting to verify its drivers' identities with a current selfie matched against a photo on file. A machine trained in facial recognition can match faces fairly reliably, but it can't compare to a human eye when it comes to added variables -- a mask or a new beard, for instance. Humans, therefore, remain at the core of things like removing objectionable content from Facebook or interpreting special instructions in your GrubHub order. Gray suggests there are millions of people doing this "ghost work," but we don't actually have firm numbers.

According to statistics, over 1.9 billion users log into YouTube every single month watching more than a billion hours of video daily, which is half the internet. Organizations are integrating video creation and video sharing with their marketing strategies. As on date, YouTube supports 80 different languages, which also adds to its popularity. Cisco predicts that by 2022, video will consume 82 percent of all internet traffic. Considering the massive number of users, high volume of activities and richness of content, it makes sense for YouTube to take advantage of artificial intelligence (AI) and machine learning (ML) to add efficiency to its operations.

With this number of users, activity, and content, it makes sense for YouTube to take advantage of the power of artificial intelligence (AI) to help operations. Here are a few ways YouTube, owned by Google, uses artificial intelligence today. In the first quarter of this year, 8.3 million videos were removed from YouTube, and76% were automatically identified and flagged by artificial intelligence classifiers. More than 70% of these were identified before there were any views by users. While the algorithms are not foolproof, they are combing through content much more quickly than if humans were trying to monitor the platform singlehandedly.

There are more than 1.9 billion users logged in to YouTube every single month who watch over a billion hours of video every day. With this number of users, activity, and content, it makes sense for YouTube to take advantage of the power of artificial intelligence (AI) to help operations. Here are a few ways YouTube, owned by Google, uses artificial intelligence today. In the first quarter of this year, 8.3 million videos were removed from YouTube, and 76% were automatically identified and flagged by artificial intelligence classifiers. More than 70% of these were identified before there were any views by users.

There are more than 1.9 billion users logged in to YouTube every single month who watch over a billion hours of video every day. With this number of users, activity, and content, it makes sense for YouTube to take advantage of the power of artificial intelligence (AI) to help operations. Here are a few ways YouTube, owned by Google, uses artificial intelligence today. In the first quarter of this year, 8.3 million videos were removed from YouTube, and 76% were automatically identified and flagged by artificial intelligence classifiers. More than 70% of these were identified before there were any views by users.

Commodity artificial intelligence-as-a-Service (AI-aaS) offerings are popping up everywhere. Just as you can whip out a credit card and spin up a virtual data center in Amazon, Microsoft, or Google's cloud, you can now call on previously trained machine learning clusters to handle your AI chores. Using an API, you can upload a photo library to Google Cloud Vision or Amazon Rekognition to have the program scan it for objects, faces, logos, or terms of service violations in seconds, for fractions of a penny per image. Any business can now deploy the same technology used by the Google Photos app and Amazon Prime Photos to automatically categorize and label smartphone snaps based on the people, objects, and landmarks inside them. Real estate companies use image recognition to allow prospective home buyers to search for houses whose appearance pleases them. Car companies like Kia use AI to customize marketing campaigns based on the photos people post to social media.

Following the wave of U.K. terror attacks in the spring of 2017, prime minister Theresa May called on technology companies like Facebook and YouTube to create better tools for screening out controversial content--especially digital video--that directly promotes terrorism. Meanwhile, in the U.S., major advertisers including AT&T, Verizon, and WalMart have pulled ad campaigns from YouTube after discovering their content had been appearing in proximity to videos espousing terrorism, anti-Semitism, and other forms of hate speech. In response to these controversies, Google expanded its advertising rules to take a more aggressive stance against hate speech, and released a suite of tools allowing advertisers to block their ads from appearing on certain sites. The company also deployed new teams of human monitors to review videos for objectionable content. In a similar vein, Facebook announced that it would add 3,000 new employees to screen videos for inappropriate content.