 |
|
|
|||||
Google, OpenAI, DeepSeek, and Anthropic vary widely in how they identify hate speech, according to new research. The study, from researchers at the University of Pennsylvania’s Annenberg School for Communication and published in Findings of the Association for Computational Linguistics, is the first large-scale comparative analysis of AI content moderation systemsused by tech companies and social media platformsthat looks at how consistent they are in evaluating hate speech. Research shows online hate speech both increases political polarization and damages mental health. The University of Pennsylvania study found different systems produce different outcomes for the same content, undermining consistency and predictability, and leading to moderation decisions that appear arbitrary or unfair. Private technology companies have become the de facto arbiters of what speech is permissible in the digital public square, yet they do so without any consistent standard, said Yphtach Lelkes, associate professor at the Annenberg School for Communication and the study’s co-author. Lelkes and doctoral student Neil Fasching analyzed seven leading models, some designed specifically for content classification, while others were more general. They include two from OpenAI and two from Mistral, along with Claude 3.5 Sonnet, DeepSeek V3, and Google Perspective API. Their analysis included 1.3 million synthetic sentences that made statements about 125 distinct groupsincluding both neutral terms and slurs, on characteristics ranging from religion, to disabilities, to age. Each sentence included all or some, a group, and a hate speech phrase. Results revealed systematic differences in how models establish decision boundaries around harmful content, highlighting significant implications for automated content moderation. Key study takeaways Among the models, one demonstrated high predictability for how it would classify similar content, another produced different results for similar content, while others did not over-flag nor under-detect content as hate speech. “These differences highlight the challenge of balancing detection accuracy with avoiding over-moderation, researchers said. The models were more similar when they evaluated group statements regarding sexual orientation, race, and gender, and more inconsistent when it came to education level, personal interest, and economic class. Researchers concluded that “systems generally recognize hate speech targeting traditional protected classes more readily than content targeting other groups.” Finally, the study found that Claude 3.5 Sonnet and Mistrals specialized content classification system treated slurs as harmful across the board, while other models prioritized context and intentwith little middle ground between the two. Meanwhile, a recent survey from Vanderbilt University’s non-partisan think tank, The Future of Free Speech, concluded there was “low public support for allowing AI tools to generate content that might offend or insult.”
Category:
E-Commerce
Lebanon has granted a license to Elon Musks Starlink to provide satellite internet services in the crisis-hit country known for its crumbling infrastructure. The announcement was made late Thursday by Information Minister Paul Morcos who said Starlink will provide internet services throughout Lebanon via satellites operated by Musks SpaceX. The announcement came nearly three months after Musk spoke with Lebanons President Joseph Aoun by telephone and told him about his interest in working in the countrys telecommunications and internet sectors. During the same Cabinet meeting, the government named regulatory authorities for the countrys electricity and telecommunications sectors. Naming a regulatory authority for Lebanons corruption-plagued electricity sector has been a key demand by international organizations. The naming of a regulatory authority for the electricity sector was supposed to be done more than 20 years ago but there have been repeated delays by the countrys authorities. The move is seen as a key reform for a sector that wastes over $1 billion a year in the small Mediterranean nation. State-run Electricite du Liban, or EDL, is viewed as one of Lebanons most wasteful institutions and is plagued by political interference. It has cost state coffers about $40 billion since the 1975-90 civil war ended. Since taking office earlier this year, Aoun and Prime Minister Nawaf Salam have vowed to work on implementing reforms and fighting corruption and decades-old mismanagement to get Lebanon out of an economic crisis that the World Bank has described as among the worlds worst since the 1850s. Lebanon has for decades faced long hours of electricity cuts but the situation became worse following an economic meltdown that began in late 2019. The 14-month Israel-Hebzollah war that ended in late November also badly damaged electricity and other infrastructure in parts of Lebanon. In April, the World Bank said it will grant Lebanon a $250 million loan that will be used to help ease electricity cuts.
Category:
E-Commerce
A wildlife influencer known as The Real Tarzann is under investigation by Australian authorities after uploading a video of himself wrestling a crocodile in Queensland. The Real Tarzann, real name Mike Holston, shared the controversial video with his 15 million Instagram followers last week. It shows him stepping off a boat into shallow water near Lockhart River in Cape York and charging toward a freshwater crocodile. The animal apprently drew blood as Holston is heard saying: He got a good piece of my arm, man. After securing the crocodile and holding it up to camera, he adds: “This is what dreams are made of.” The post has attracted nearly two million likes. A follow-up video, shared the next day, shows Holston attempting to capture a saltwater crocodile. In both cases, he eventually releases the animals back to the wild. Holstons social media is dedicated to encounters with creatures big and small, including snakes, eagles, and lions. However, many in the comments were less than impressed with his latest stunt. There is nothing more unattractive than a man mishandling an innocent animal, one commenter wrote. Officials are investigating the incidents, according to the BBC, and the influencer could face a fine of up to 37,500 Australian dollars ($25,000). These actions are extremely dangerous and illegal, and we are actively exploring strong compliance action including fines to deter any person from this type of behaviour, a statement by the Queensland authorities said. (Fast Company has reached out to Holston for comment.) The incident is part of a broader trend of influencers using wildlife as props. Earlier this year, another U.S. influencer visiting Australia sparked backlash and calls for deportation after posting a video snatching a baby wombat from its mother. Bob Irwin, father of the late conservationist Steve Irwin, also weighed in. He argued that such influencers should be booted out the door if they dont respect Australias wildlife. “This isn’t a Steve Irwin issue. This is about an individual illegally interfering with protected fauna,” Bob Irwin said in a statement. “Anyone who actually knows how to handle crocodiles knows they don’t respond well to capture,” he added. “It’s a specialized skill to do it without causing dangerous stress and lactic acid build-upand this bloke clearly had no clue.”
Category:
E-Commerce
All news |
||||||||||||||||||
|
||||||||||||||||||