Sam Altman, co-founder and CEO of OpenAI, says the company knows how to build artificial general intelligence, anticipates AI agents entering the workforce in 2025, and is now focusing on how to achieve artificial superintelligence.
Altman shared thoughts on the company’s progress in a blog post, reflecting on events that challenged OpenAI’s governance model and required structural changes. He went into new detail on the period when he was briefly removed from the organization, calling it a failure of oversight that tested core leadership principles. He also acknowledged individuals who worked behind the scenes to stabilize operations, emphasizing how these developments reaffirmed the need for a framework suited to evolving technology and high-capital demands.
Industry skepticism of AGI in 2025
OpenAI’s claim that it knows how to achieve AGI, traditionally defined, has drawn criticism and praise. Some experts point to surveys indicating only a 50% chance of high-level machine intelligence closer to the 2050s, as InformationWeek reported. Others question whether breakthroughs in autonomous learning and fully transparent reasoning can materialize so soon. Altman’s timeline for AI agents joining the workforce has stirred debate, though Salesforce is already pushing the concept with its Agentforce product. He argues that iterative releases and direct user feedback will improve safety while pushing functionality forward.
According to Altman, OpenAI’s ambitions extend beyond AGI toward superintelligence, a step that some observers see as inevitable for advanced AI systems. Altman wrote that the prospect of superintelligent tools could trigger far-reaching discoveries in science and engineering. He believes cautious deployment and alignment research can address ethical considerations. Many wonder how rapidly these technologies will influence global industries, including crypto markets, where AI-driven automation and analytics are already making waves with current technology.
Thus, skepticism persists around claims of near-term AGI. Critics highlight challenges in enabling genuine understanding, cross-domain reasoning, and context awareness. Altman maintains that each model update, combined with user interactions, helps refine output and guides safety mechanisms. I personally predicted AGI would not be achieved in 2025 amid other New Year predictions as the gulf between current models and AGI seems too broad to cover in 12 months.
Altman also addressed OpenAI’s shift from pure research to a product-oriented approach. He noted that original assumptions about capital requirements and theoretical objectives needed revisiting once user adoption accelerated. He says Iteration reflects OpenAI’s commitment to letting society adapt to emerging tools without halting development.
Altman concluded that OpenAI’s mission remains consistent despite recent turmoil. He believes superintelligence will eventually augment human potential and tackle complex challenges. He added that wide adoption and community feedback will help shape practical safeguards, with many in the AI sector closely watching OpenAI’s progress.