September 10, 2025 News
Thinking Machines Lab Develops Method to Make AI Models Generate Reproducible Responses
Mira Murati's Thinking Machines Lab published research addressing the non-deterministic nature of AI models, proposing a solution to make responses more consistent and reproducible. The approach involves controlling GPU kernel orchestration during inference processing to eliminate randomness in AI outputs. The lab suggests this could improve reinforcement learning training and plans to customize AI models for businesses while committing to open research practices.
Skynet Chance (-0.08%): Making AI models more deterministic and predictable reduces one source of unpredictability that could contribute to AI safety risks. More consistent AI behavior makes systems easier to control and understand, slightly reducing alignment concerns.
Skynet Date (+0 days): While this improves AI reliability, it doesn't fundamentally accelerate or decelerate the timeline toward potential AI control problems. The research addresses technical consistency rather than capability advancement that would change risk timelines.
AGI Progress (+0.03%): Improved determinism and enhanced reinforcement learning efficiency represent meaningful technical progress toward more reliable AI systems. Better RL training could accelerate development of more capable and controllable AI models.
AGI Date (+0 days): More efficient reinforcement learning training and reproducible responses could modestly accelerate AGI development by making AI training processes more reliable and effective. However, this addresses training efficiency rather than fundamental capability breakthroughs.
OpenAI Signs Massive $300 Billion Cloud Computing Deal with Oracle
OpenAI has reportedly signed a historic $300 billion cloud computing contract with Oracle spanning five years, starting in 2027. This deal is part of OpenAI's strategy to diversify away from Microsoft Azure and secure massive compute resources, coinciding with the $500 billion Stargate Project involving OpenAI, SoftBank, and Oracle.
Skynet Chance (+0.04%): Massive compute scaling could enable more powerful AI systems that are harder to control or monitor. The diversification across multiple cloud providers also creates a more distributed infrastructure that could be more difficult to govern centrally.
Skynet Date (-1 days): The enormous compute investment accelerates AI capability development timeline significantly. Starting in 2027, this level of computational resources could enable rapid advancement toward more powerful AI systems.
AGI Progress (+0.04%): Access to $300 billion worth of compute power represents a massive scaling of resources that directly enables training larger, more capable AI models. This level of computational investment is a significant step toward the compute requirements needed for AGI.
AGI Date (-1 days): The massive compute contract starting in 2027 substantially accelerates the timeline for AGI development. This level of computational resources removes a key bottleneck and enables OpenAI to pursue much more ambitious AI training projects.
TwinMind Raises $6M for AI-Powered "Second Brain" App That Continuously Listens and Transcribes Speech
Former Google X scientists have launched TwinMind, an AI app that runs continuously in the background to capture ambient speech and build a personal knowledge graph. The startup raised $5.7 million in seed funding and released their Ear-3 speech model supporting 140+ languages with 5.26% word error rate. The app processes audio on-device for privacy, can run 16-17 hours without battery drain, and has attracted over 30,000 users.
Skynet Chance (+0.04%): The always-listening AI assistant represents a step toward pervasive AI monitoring and data collection, though privacy measures like on-device processing and audio deletion partially mitigate immediate control risks. The technology normalizes constant AI surveillance of human conversations and activities.
Skynet Date (+0 days): The widespread deployment of ambient AI systems that continuously monitor human behavior could accelerate the timeline by normalizing pervasive AI presence. However, the focus on privacy-preserving, on-device processing doesn't significantly change the overall pace toward concerning AI capabilities.
AGI Progress (+0.03%): The development demonstrates progress in multimodal AI systems that can understand context across speech, vision, and web browsing simultaneously. The ability to build personalized knowledge graphs from continuous real-world interaction represents advancement toward more contextually aware AI systems.
AGI Date (+0 days): The successful deployment of always-on, context-aware AI systems with efficient on-device processing suggests faster progress in creating AI that can understand and interact with human environments continuously. The commercial success and user adoption indicates viable pathways for pervasive AI integration.