South Korean director Yeon Sang-ho continues to explore artificial intelligence in his new psychological thriller, Paradise Lost. The film stars Kim Hyun-joo as a mother who uses augmented reality glasses to interact with an AI version of her kidnapped son, frozen at age 10. When the real son returns as an adult, she must choose between the digital recreation and her living child. Paradise Lost follows Yeon's earlier AI-themed works, including the 2023 Netflix film JUNG_E and this year's Colony.
Jeff Dean, who spent 27 years at Google before departing, spoke with Stanford professor Fei-Fei Li about the limits of AI self-improvement. Dean argued that the idea of infinitely fast improvement loops is not realistic. His new company currently runs about 20 automated research loops and aims to boost expert productivity by 20 to 50 times rather than replacing human workers entirely.
Seattle's Fire Department has faced criticism for using AI to analyze 911 medical calls since 2023 without informing callers. The city's chief privacy officer raised equity concerns in 2019, but those warnings went largely unheeded. The AI system sometimes routes patients to a nurse line in Texas instead of dispatching ambulances, causing delays. City officials now plan to publicly address AI use in early 2027.
Key Takeaways
- Director Yeon Sang-ho released Paradise Lost, a thriller about a mother choosing between an AI recreation of her son and the real child
- Jeff Dean left Google after 27 years and now runs a company with about 20 automated AI research loops
- Seattle used AI on 911 medical calls since 2023 without caller knowledge despite prior privacy warnings
- Democrats won the midterm elections, and voters passed 43 of 44 local data center moratoria
- Japan is preparing a cybersecurity package and has approached OpenAI and Anthropic for broader model access
- Hobbyists use GPT-6 Astra for vibe manufacturing, turning spoken descriptions into physical parts, with one user losing about $2,000 on failed pieces
- OpenMatter offers decentralized computing as an alternative to large data centers after election backlash
- Mercedes began rolling out Liquid AI models in North American cars in October for conversational features
- Consumer AI chatbots are becoming a healthcare front door, raising concerns about unregulated clinical advice
- Nonprofits are urged to use AI for efficiency but must always have humans verify outputs like grant writing
South Korean director Yeon Sang-ho explores AI in new thriller Paradise Lost
Director Yeon Sang-ho has made another film about artificial intelligence with Paradise Lost. The psychological thriller stars Kim Hyun-joo as a mother who uses an AI version of her missing son through augmented reality glasses. When the real son returns as an adult, she struggles to choose between the two. Yeon's earlier AI films include the 2023 Netflix film JUNG_E and this year's Colony.
Director Yeon Sang-ho revisits AI themes in thriller Paradise Lost
South Korean director Yeon Sang-ho released Paradise Lost, a psychological thriller about AI. Kim Hyun-joo plays a mother whose son was kidnapped during a school trip. She turns to an AI recreation of her son that stays 10 years old. When the real son returns as a grown man played by Bae Hyeon-seong, she faces a difficult choice. The film is the latest from the Train to Busan director.
Jeff Dean says self-improvement AI loops cannot run infinitely fast
Jeff Dean, who left Google after 27 years, co-founded a company building AI that can drive scientific discovery. Speaking with Stanford professor Fei-Fei Li, Dean said the idea of running improvement loops infinitely fast is implausible. His company currently runs about 20 automated loops focused on machine learning research and engineering. The goal is to boost expert productivity by 20 to 50 times, not to replace humans.
Seattle ignored warnings about AI use on 911 medical calls
Seattle's Fire Department has used AI to help analyze 911 medical calls since 2023 without telling callers. In 2019, the city's chief privacy officer warned the department about equity risks from AI, but her concerns were largely ignored. The AI system routes some patients to a Texas nurse line instead of sending immediate help, which has caused ambulance delays. City officials promise a public update on AI use in early 2027.
AI will not save nonprofits but could help them work smarter
A Nonprofit Quarterly article says AI will not rescue nonprofits but could help them operate more efficiently. The piece acknowledges real fears about AI including job loss, hallucinations, bias, and environmental costs. It stresses that nonprofits using AI for grant writing or reports must always have humans check the output for accuracy. The author urges nonprofits to move past abstract discussions and have honest conversations about practical AI use.
Blue wave election threatens AI industry amid backlash
Democrats won control of the House and Senate in midterm elections, creating a new political threat for AI companies. Voters passed 43 out of 44 local data center moratoria, showing strong opposition to data center construction. Experts say the results likely end Republican efforts to block state-level AI regulations. Leaders from groups like the Guardrails Alliance and Future of Life Institute say voters are tired of having no say in how AI affects their lives. The backlash is described as bipartisan and focused on rising costs and community impacts.
Japan pressures US tech giants on security and AI access
Japan is preparing an emergency cybersecurity package that could be discussed as early as Friday, Nov. 6. Officials want earlier notice of software security weaknesses from major US tech and AI companies. Prime Minister Sanae Takaichi ordered stronger protections after a rise in cyberattacks in October. Japan also plans to use AI to simulate cyberattacks and find system weaknesses. The government has approached OpenAI and Anthropic about broader access to their AI models following suspected AI-assisted cyberattacks.
People use AI to build real-world products with GPT-6 Astra
Hobbyists and developers are using GPT-6 Astra to turn spoken descriptions into real parts using fabrication tools. Users create items like robot arms, brain models from MRI data, children's toys, and answering machines. Physical design errors can cost real money, with one user spending about $2,000 on failed parts. The process, called vibe manufacturing, works for prototypes but still requires careful checking for safety and quality. Makers still need to verify tolerances, materials, and regulatory compliance before producing certified products.
OpenMatter offers decentralized computing as data center alternative
OpenMatter Network says the midterm election results show voters oppose massive data centers due to energy use, environmental impact, and rising costs. The company offers a decentralized computing model that spreads computing resources across many locations instead of concentrating them in huge facilities. Its platform turns widely available hardware into network resources using algorithms and verifiable computing. Leaders say this approach can support enterprise AI workloads while reducing pressure on local communities. The model also gives organizations more control over data sovereignty and locally managed resources.
Mercedes adds Liquid AI models to cars in North America
Mercedes has started rolling out Liquid AI models in some of its cars in North America beginning in October. The luxury carmaker focuses on letting drivers talk to their cars to adjust features like heating and infotainment rather than controlling driving directly. Mercedes says this AI capability could help attract customers in a competitive market. The move represents a new step in bringing conversational AI into everyday vehicle experiences.
Consumer AI Is Healthcare's New Front Door. Who's Deciding What Happens There?
Consumer AI chatbots are becoming a first point of contact for healthcare, but they often blur the line between providing information and making clinical judgments. A patient might receive a suggestion of pneumonia from a chatbot, but it cannot perform the physical exam needed for a real diagnosis. This unregulated use of AI outside of healthcare governance structures creates risks, such as patients skipping necessary provider visits based on AI advice. Health systems and builders must define clear boundaries to ensure these tools act as assistants rather than replacements for professional care.
Sources
- South Korean director Yeon Sang-ho revisits AI themes in latest thriller 'Paradise Lost'
- South Korean Director Yeon Sang-Ho Revisits AI Themes in Latest Thriller 'Paradise Lost'
- “Implausible” That Self-Improvement Loops Will Run Infinitely Fast: Jeff Dean
- Internal report raised concerns about AI use on Seattle 911 calls, went ignored
- AI won’t save nonprofits, but it might free them up to save themselves | Nonprofit Quarterly | Civic News. Empowering Nonprofits. Advancing Justice.
- ‘Total repudiation’: The blue wave is about to hit AI
- Japan Steps Up Pressure on US Tech Giants Over Security and AI
- Forget Vibe Coding – People Are Now Using AI to Build Real-World Products
- OpenMatter Network Positions Decentralized Computing as an Alternative to Data Centers for AI
- Exclusive | Mercedes Gives Its Cars an AI Upgrade
- Consumer AI Is Healthcare's New Front Door. Who's Deciding What Happens There?
Comments
Please log in to post a comment.