Meta Bans Political Advertisers from Using Generative AI Advertising Tools


The Shanghai Cyberspace Administration launched a special law enforcement campaign targeting the 'misuse of AI', aiming to address illegal activities related to generative AI. The operation focused on issues such as AI face-swapping and voice-changing, which pose threats to personal privacy and disrupt the online ecosystem. During the enforcement, some companies were found not to have developed and used AI technology in accordance with regulations, and it was emphasized that strengthening supervision is needed to prevent risks.
Meta's DreamGym framework, developed with universities, uses simulated reinforcement learning to reduce training costs and improve feedback reliability for large language models by dynamically adjusting task difficulty.....
Japanese AI startup Sakana AI secured a $2 billion Series B financing, with a post-money valuation of $26.5 billion. The round was led by Mitsubishi UFJ Financial Group, with participation from several new investors. The company was founded by former Google researchers and focuses on developing lightweight AI models based on small data, native to Japanese and aligned with Japanese culture. The funding will be used for research and development, team expansion, and other purposes, with plans to achieve business goals by 2026.
Yann LeCun criticizes big tech's focus on large language models as a strategic error, arguing they lack true intelligence, world understanding, and reasoning. He advocates for 'world models' as the next breakthrough.....
Google denies liability for AI 'hallucinations' in a defamation lawsuit by anti-corporate diversity activist Robbie Starbuck, who claims its AI falsely linked him to sexual assault and white supremacy. He previously sued Meta for similar AI errors.....