OpenAI claims GPT-6 Astra is an ethereal 'Alien Mind' with AGI-like qualities — company warns of alignment challenges as new frontier leader emerges
OpenAI's GPT-6 Astra is announced as a new frontier model, with claims of AGI from Jensen Huang and a blog post from chief scientist Jakub Pachocki describing it as an 'alien mind'. The article discusses the model's benchmarks, training on 100,000 Blackwell GPUs, and its cheaper per-task cost compared to Anthropic's Fable 5.1. It also covers Pachocki's caution about alignment, calls for international coordination, and the context of rising competition from cheaper alternatives like Deepseek and Google Gemini.
Related Tech News
AI CuratedSamsung and SK hynix Memory Inventories Crash Below 10 Days as AI Demand Triggers Historic Shortage, Says Report
Samsung and SK hynix are facing a historic memory shortage, with inventories reportedly down to around ten days, according to Business Korea. AI data center demand is straining production, and the launch of capacity-intensive HBM4 products in 2027 could worsen the crunch. The report highlights growing supply pressures across the memory market.
Big Tech eyes glacier-strewn Patagonia for building mega AI data centers — region offers 17,300 glaciers, coldness, and cheap energy, but fiber lines are scarce
A Reuters report highlights Patagonia's potential as a data center hub, with interest from Big Tech like Amazon, Google, and OpenAI. The region offers cold climate and abundant energy (hydroelectric, wind, shale gas), but lacks fiber connectivity and faces political uncertainty with upcoming elections. OpenAI has confirmed a $25B, 500MW data center in Argentina, while other companies like Green Capital and FlexDomes plan facilities. Structural weaknesses in power grid and connectivity add costs.
More than 1 in 10 chance AI ‘could kill all humans,’ says Anthropic safety lead after colleague quits
News that an Anthropic safety lead estimated a greater-than-10% chance AI could kill all humans by 2030, following a colleague's resignation over what he called 'lax' safety and a race to superintelligence.
More than 10% chance AI 'could kill all humans' in the next 10 years, Anthropic safety researcher says — departing employee says AI companies are 'gambling with our lives'
Tom's Hardware reports on warnings from Anthropic safety researcher Evan Hubinger and departing employee Jacob Coxon, who state that AI companies are racing to superintelligence without adequate safety measures, with Hubinger estimating a >10% chance of human extinction within ten years.