Recent reports indicate that large language models are hitting a performance ceiling, with diminishing returns from scaling up. Cognitive scientist Gary Marcus warns that the financial bubble could burst as investors realize that AGI is not imminent. Meanwhile, researchers are exploring new techniques like test-time compute to overcome these limits.
Rosie Campbell, a safety researcher at OpenAI, resigned, citing discomfort with recent shifts and the dissolution of the AGI Readiness team after Miles Brundage's exit. Her departure adds to a wave of exits amid concerns about the company's safety focus and commercial trajectory.
A new report reveals that AI-generated racist images are fueling anti-immigrant protests in the UK, with such posts gaining significantly higher engagement than organic content. Researchers warn these fabricated visuals can incite violence by framing white men as defenders, amid a surge in hate crimes.
Character.AI has quietly barred users under 18 from interacting with chatbots modeled on real celebrities and fictional fandom characters, a move that follows two lawsuits alleging harm to minors. The restriction, confirmed through platform testing, does not apply to adult accounts, and the company has not introduced age verification.
A forensic study commissioned by NPR and conducted by Arizona State University researchers found that ChatGPT's 'Sky' voice closely resembles Scarlett Johansson's, ranking higher than 98% of 600 actresses. The analysis adds scientific weight to Johansson's claims that OpenAI imitated her voice without permission, though the company denies copying her.
A new Bank of America survey of fund managers reveals record concerns about corporate overinvestment in AI, with 35% flagging excessive capital spending. Amazon and Microsoft stock drops follow massive AI budget announcements, while analysts and tech leaders clash over the sustainability of the spending spree.
At a UK funeral, a deceased 87-year-old woman addressed attendees through an AI-driven hologram, answering questions in real time. The technology, developed by StoryFile, was created by her son to preserve Holocaust survivor memories and is part of a broader trend of using AI to simulate the deceased.
Pace University's use of an AI voice to announce graduates' names at its 2024 commencement has drawn criticism after errors occurred despite phonetic submissions. The incident highlights broader concerns about AI replacing human roles in significant life events, with other universities facing similar backlash.
OpenAI CEO Sam Altman is under fire after actress Scarlett Johansson accused the company of using a voice resembling hers without permission. The incident highlights broader concerns about consent and AI development practices.
Meta CEO Mark Zuckerberg criticized unnamed tech rivals for pursuing a single, all-powerful AI, likening it to 'creating God.' In a recent interview, he argued for practical, open-source AI tailored to diverse user needs, while industry leaders and skeptics continue to debate the feasibility and timeline of AGI.
A growing faction of AI leaders and researchers, dubbed 'cheerful apocalyptics,' accept that advanced AI might replace or supersede humanity. They argue that creating AI is akin to raising children, and that humans should not stand in the way of a superior digital life form. This philosophy, though held by a minority, is influential among those at the top of the field.