Special Projects (2016)

Commenters debated OpenAI’s Special Projects in terms of long-term vision and the concrete ambitions described there - particularly the idea of huge simulations of many long‑lived agents. samayashar and Pranav_Ghoghari praised sticking to a decade‑long vision and even projected robotics as the next trillion‑dollar frontier, while codeulike connected the simulation idea to Greg Egan’s Crystal Nights and jmathai highlighted the goal of building agents that can write programs, noting Anthropic’s early lead. goldenbrillianc and others reflected on how unexpected it was that general‑purpose models trained on massive internet data would outperform traditional, narrowly engineered program generators. ronsor and tehmillhouse pointed out leadership departures and reported deep disagreements about company direction.

Opinion split sharply over causes of progress and OpenAI’s institutional choices. bonoboTP, mapBasketWand and willy_k emphasized technical inflection points - GPGPU, data and compute growth, and the 2017 transformer paper - as the enablers of LLMs, while legulere warned against retrospective bias and owebmaster reminded readers that “AI” long predates LLMs. thatsabadlook criticized OpenAI’s move toward profit and secrecy as inconsistent with original openness, whereas AdamN credited sustained funder engagement for keeping projects alive. The debate centers on whether success is primarily vision‑driven or the product of funding, infrastructure and cumulative technical shifts, and whether corporate strategy aligns with the founders’ original ideals.