Sam Altman Says AI Agents Will Transform the Workforce in 2025

Sam Altman says the two years since the launch of ChatGPT, a period that has catapulted him to fame as the public face of the artificial intelligence industry, have been the most “unpleasant years of my life so far.” In a new blog post, Altman reflected on the path that he’s walked since ChatGPT’s November 2022 start, including what he learned from his very-public firing in 2023, and made a big prediction about AI’s impact in 2025. Here are the biggest takeaways from Altman’s January 2025 lengthy blog post, titled “Reflections.” Altman’s firing still haunts him Altman was publicly fired by OpenAI’s board in November 2023, just before ChatGPT’s first birthday. Five days later, he was reinstated as CEO. In the blog post, Altman reveals some personal details regarding the firing, which happened over a video call while he was in Las Vegas. Looking back, he says the whole event was a “big failure of governance by well-meaning people, myself included,” but one that he believes has made him a more thoughtful leader. Another lesson from the firing? The importance of having a board with diverse viewpoints and experience handling unexpected challenges. In particular, Altman singled out two figures who he said “went so far above and beyond the call of duty” to rescue Altman from his brief banishment: Airbnb founder Brian Chesky and venture capitalist Ron Conway. Without going into detail, Altman recalled “being in the foxhole” with Chesky and Conway, who “used their vast networks for everything needed and were able to navigate many complex situations. And I’m sure they did a lot of things I don’t know about.” He believes in scaling laws Altman has long been a believer in scaling laws, a mathematical assumption that the more data a neural network is trained on, the smarter it becomes. In his blog post, he theorized that businesses also have a scaling law: As growth increases, so does turnover. Acknowledging that OpenAI’s executive team has seen a massive amount of turnover since ChatGPT’s launch, Altman wrote that “startups usually see a lot of turnover at each new major level of scale, and at OpenAI numbers go up by orders of magnitude every few months.” According to Altman, the fracturing of OpenAI’s C-suite, including the departures of chief technology officer Mira Murati and chief scientist Ilya Sutskever, are a natural result of OpenAI’s ascendancy. OpenAI’s structure, and its future OpenAI’s leadership reportedly spent much of 2024 determining how to transform its current structure as an entity with a capped for-profit arm and a nonprofit arm into a more conventional moneymaking entity. Altman wrote in his post that he had “no idea we would need such a crazy amount of capital” to develop super-advanced artificial intelligence. To obtain that kind of capital, OpenAI is planning on converting its for-profit arm into a public benefit corporation. In an official statement released in December, OpenAI wrote that “investors want to back us but, at this scale of capital, need conventional equity and less structural bespokeness.” A key test of this planned new OpenAI structure will be how the company sells enterprises on AI agents, which are designed to take specific actions and automate workflows. Altman wrote in his blog that 2025 could be the year that AI agents are integrated into the workforce and predicted they would “materially change the output of companies.” Beyond 2025, OpenAI is turning its aim beyond useful tools to “superintelligence,” super-advanced AI models capable of outperforming humans at nearly any task and ushering in a new era of abundance and prosperity. “We love our current products,” wrote Altman, “but we are here for the glorious future.” OpenAI’s naming struggles Altman is effusive about OpenAI’s capabilities in nearly all areas, save for one notable exception: naming stuff. The company has a history of giving its new AI models and products confusing names like GPT-4, GPT-4o, GPT-4o Mini, o1, and o1 Mini. In July 2024, when announcing GPT-4o Mini, Altman responded to a post on X suggesting that OpenAI needed to revamp its naming scheme with “lol yes we do.” In his blog post, Altman says that originally, ChatGPT was named Chat With GPT-3.5, adding that OpenAI is “much better at research than we are at naming things.” All together, the post is nearly 2,000 words, so if you don’t feel like reading the whole item, you’re in luck: When asked to summarize the screed in a single sentence, ChatGPT 4o provided the following: “OpenAI’s journey over the past nine years, marked by the launch of ChatGPT and transformative progress in AI development, has been a mix of extraordinary innovation, intense challenges, and a vision for creating beneficial AGI, culminating in a reflection on resilience, gratitude, and the promise of a super-intelligent future.” BY BEN SHERRY @BENLUCASSHERRY