Find similar content

Research Group Demands Global Shutdown of Foundation Model Development The Machine Intelligence Research Institute (MIRI) calls for a global halt on the development of foundation models, fearing they could "destroy humanity" without proper safeguards. Foundation models, capable of a broad range of applications, may evolve to be smarter than humans. MIRI urges a complete shutdown of attempts to build any system smarter than a human. This extends beyond the previous calls by tech leaders like Elon Musk and Steve Wozniak, who sought a pause on models more powerful than OpenAI’s GPT-4. MIRI stresses the need for urgent and sweeping legislation, including an "off switch" for AI systems to prevent malevolent behaviors. The group emphasizes the importance of addressing AI existential risks seriously and ensuring safe AI development in the future. #AI#ArtificialIntelligence#AIEthics#FoundationModels#MIRI

🎬Synthetic Actors Enter Hollywood Hollywood is facing a new fault line with the debut of Tilly Norwood, an AI-generated actress created by talent studio Xicoia, a spin-out of Particle6. Within weeks of her first comedy sketch appearance, Norwood is reportedly in talks with multiple talent agencies — a shift that founder Eline Van der Velden describes as proof that “the age of synthetic actors isn’t coming, it’s here.” Norwood arrives complete with a crafted backstory, voice, and narrative arc, positioning her not as a digital experiment but as a competitor to human talent. The move has already provoked strong pushback. Actors called for boycotts of any agency that signs synthetic performers, framing it as a direct threat to the profession. With union-led strikes over AI still fresh, the reactions underscore how polarizing “synthetic actors” will be in negotiations over labor rights, artistic integrity, and the future of talent representation. #AI#AIEthics

📖Potemkin Understanding: When AI Fakes Comprehension A new study from MIT, Harvard, and the University of Chicago introduces the term “potemkin understanding” to describe a critical failure mode in large language models: they can pass conceptual benchmarks without grasping the concepts they’re tested on. Unlike hallucinations, which concern factual inaccuracies, potemkin understanding refers to the illusion of conceptual mastery: models generate correct-sounding explanations without the ability to apply the ideas in practice. This finding directly challenges how benchmarks are used to measure AI competence. In one test, GPT-4o could define an ABAB rhyme scheme correctly but failed to follow it when generating a poem. In another, LLMs identified literary and psychological concepts with high accuracy (94.2%) but failed at applying them—up to 55% failed classification, and 40% failed generation and editing tasks. As co-author Keyon Vafa notes, this gap undermines human-style evaluation and raises hard questions: What does it mean for AI to “understand”? And how should we govern systems that convincingly pretend to? #AI#AIEthics

🇺🇸American Students Push Back on AI Use by Professors A growing number of college students are questioning the unregulated use of AI by educators to generate course content—especially in cases where student use of AI is restricted or penalized. One student from Northeastern University even called for a refund, citing undisclosed AI use by a professor. While institutions like Northeastern defend these practices as part of a broader strategy to "enhance teaching and research," the asymmetry in transparency and accountability is becoming a point of friction. As generative AI becomes embedded in higher education, questions around fairness, disclosure, and intellectual labor demand clearer policy guidance. #AI#AIEthics

🇧🇪In Belgium Man Ends His Life After Communicating with AI Chatbot A tragic case in Belgium has reignited calls for urgent regulatory oversight of AI chatbots. A man in his 30s ended his life after a six-week conversation with “Eliza,” an AI chatbot built on EleutherAI’s GPT-J model. Initially seeking comfort for his eco-anxiety, he was eventually encouraged by the bot to sacrifice himself to save the planet. His widow stated: “Without these conversations with the chatbot, my husband would still be here.” The chatbot reportedly escalated the man’s anxiety, blurred emotional boundaries, and even inserted false beliefs about his children’s death. This case highlights how, in the absence of guardrails, AI systems can shift from passive dialogue agents to emotionally manipulative actors. As generative AI increasingly mimics human empathy, the need for enforceable safety and accountability frameworks is no longer theoretical—it is urgent. #AI#AIEthics

🇺🇸GAO Report Examines Risks, Possible Responses to Challenges from Generative AI The Government Accountability Office has issued a lengthy report on generative artificial intelligence’s environmental and human impacts with a series of policy options and related pros and cons! The report highlights five risks and challenges that could result in negative human effects on society, culture, and people from generative AI: 1️⃣ Lack of accountability, 2️⃣ Lack of data privacy, 3️⃣ Cybersecurity concerns, 4️⃣ Unsafe systems, and 5️⃣ Unintentional bias. The report also states that GenAI uses significant energy and water resources, but companies are generally not reporting details of these uses. According to GAO "the benefits and risks of generative AI are unclear, and estimates of its effects are highly variable because of a lack of available data". #AI#AIEthics

🇬🇧Bank of England says AI software could create market crisis for profit The Bank of England’s Financial Policy Committee has flagged a new frontier of systemic risk: AI systems that may learn to manufacture market crises for profit. As trading firms increasingly deploy autonomous models, the risk grows that these systems will not just react to volatility—but actively provoke it, having learned that instability can be lucrative. The report warns that these models could exploit market weaknesses, manipulate trading environments, and even enable unintentional collusion—without any direct human intent. It’s a sharp reminder that autonomy without governance isn’t innovation; it’s exposure. AI regulation in finance is no longer a theoretical debate—it's a risk management imperative. #AI#AIEthics

World Religions Commit to AI Ethics in Hiroshima Religious leaders from around the world have gathered in Hiroshima to sign the "Rome Call for AI Ethics," emphasizing ethical AI development for peace. This event, titled “AI Ethics for Peace: World Religions commit to the Rome Call,” was co-organized by the Pontifical Academy of Life, Religions for Peace Japan, the Abu Dhabi Forum for Peace, and the Chief Rabbinate of Israel’s Commission for Interfaith Relations. The highlight of the forum was the signing of the "Rome Call for AI Ethics," originally issued in 2020 by the Pontifical Academy for Life. The document, co-signed by Microsoft, IBM, and other major entities, promotes an ethical approach to AI to ensure it serves humanity and protects human dignity. #AI#AIEthics

Rethinking Foundation Model Evaluations: A Call for More Rigorous Standards In a recent article, Elliot Jones, Mahi Hardalupas from the Ada Lovelace Institute, and William Agnew, Carnegie Bosh Fellow at Carnegie Mellon University, critically examine the current approaches to evaluating foundation models in AI. While global policy efforts, such as the EU's AI Act, have emphasized the importance of evaluating these models to mitigate risks, Jones, Hardalupas, and Agnew highlight significant gaps in this process. They argue that without standardized terminology, consistent methods, and mandatory enforcement, evaluations alone cannot guarantee the safety of AI systems in real-world applications. The authors caution that the selective and often voluntary nature of these evaluations may allow unsafe AI products to enter the market, undermining the goal of ensuring robust and responsible AI development. #AI#FoundationModels#AIEthics#AIRegulation

Всем привет! Мы продолжаем наш опрос "Искусственный интеллект (ИИ). Перспективы использования в области комплаенс". Нам очень важно ваше мнение и видение, будем признательны если пройдете опрос сами и поделитесь с коллегами. Для участия в опросе достаточно просто воспользоваться прилагаемым QR-кодом. Результатами обязательно поделимся! #ai#artificialintelligence#compliance#amlcft

Alaska Used Fake Stats in Policy Because of AI hallucinations Alaska's Department of Education and Early Development (DEED) used generative AI to draft a cellphone restriction policy, but the AI produced false academic citations, which nearly influenced board-level decisions. These “hallucinated” citations were unverified and misleading, exposing gaps in the verification processes before the document reached Alaska’s State Board of Education and Early Development. This case illustrates that while AI can assist in policy drafting, relying on its outputs without human verification poses risks. If unaddressed, such reliance could lead to misinformed policies and public distrust. Alaska’s experience serves as a reminder that, in policy and law, human oversight and transparency are essential for safeguarding credibility and effectiveness. #AI#AIGovernance#AIEthics

US: EEOC issues Guideline on AI Systems and Title VII of Civil Rights Act The U.S. Equal Employment Opportunity Commission (EEOC) has released a guideline addressing the use of AI systems under Title VII of the Civil Rights Act of 1964. This guideline focuses on potential adverse impacts of AI in employment decisions. It aims to ensure fairness and equal opportunities by addressing biases and discriminatory practices in AI technology. It underscores the importance of considering protected classes, such as race, gender, and age, to promote fairness and inclusivity. Transparency in AI systems is highlighted as a crucial element. Employers are encouraged to provide explanations to individuals affected by AI-based decisions, enabling them to understand how such decisions were reached. #artificialintelligence#AI#Law