5 interesting quotes from Sam Altman's livestream (www.businessinsider.com)

OpenAI CEO Sam Altman used a livestream Q&A to lay out ambitious infrastructure and product goals while clarifying policy stances. He said OpenAI aspires to build an “infrastructure factory” that could produce 1 gigawatt of compute per week (on top of reported commitments for roughly 30 GW), with a target to reduce lifecycle costs to about $20 billion for that equipment over five years. For scale context he referenced NVIDIA’s 10 GW ≈ 4–5 million GPUs metric. On research automation, Altman gave concrete timelines: an “automated AI research intern” by September 2026 and a “true automated AI researcher” by March 2028 — signals that OpenAI is prioritizing automated model development as a path toward AGI, now framed as measurable milestones rather than abstract definitions. He also noted the company’s completed restructuring and its definitive agreement with Microsoft, which colors how and when OpenAI will declare major breakthroughs. Altman addressed customer-facing tradeoffs and safety: he apologized for a poorly chosen erotica example while reiterating a commitment to “treat adult users like adults” within safety limits (no enabling of clearly harmful activity), and warned there won’t be carte blanche even with liability waivers. He acknowledged user attachment to older models (e.g., 4o), saying there’s no immediate plan to sunset them but not promising perpetual availability. Taken together, the comments signal pushes for massive compute scale and automated research workflows, balanced with cautious, product-level safety and retention decisions that will shape deployment choices and governance in the near term.