The Stress of Working with Artificial Intelligence

OpenAI CEO Sam Altman recently discussed the challenges and pressures that come with working in the field of artificial intelligence (AI). Altman, who founded OpenAI in 2015, believes that as the world moves closer to achieving powerful AI capabilities, people's stress levels will rise dramatically.

Altman described working with AI as a "very stressful thing," citing the high stakes involved in developing artificial general intelligence (AGI). He believes that these high stakes have driven some individuals to the point of madness. Altman expects that as the world comes closer to achieving powerful AI, we will witness more "strange things" happening globally.

During a discussion at the World Economic Forum in Davos, Altman highlighted the importance of being prepared for the challenges associated with AGI. He referenced a board shakeup at OpenAI, which saw him temporarily removed as CEO before being reinstated. Altman saw this experience as a microcosm of the increased tension and stress that will occur as AGI becomes a reality.

Reflecting on the shakeup, Altman recognized the need for better preparation and handling of looming issues within OpenAI. He stressed the importance of not leaving important but non-urgent problems unresolved. Altman admitted that the board had become too small and lacked sufficient experience, but due to the tumultuous events of the past year, these issues were neglected. He emphasized the necessity of higher levels of preparation, resilience, and thinking about all possible ways things can go wrong.

As we continue to make progress in developing powerful AI, it is crucial to be aware of and address the potential challenges that lie ahead. Altman's experiences and insights serve as a reminder of the importance of preparedness and careful consideration in navigating this rapidly evolving field.

Technology in a Turbulent World: OpenAI's Perspective

Altman expressed his surprise at the New York Times' decision to sue OpenAI since, according to him, the two entities had been engaged in productive negotiations. He emphasized that OpenAI had actually intended to compensate the publisher generously.

Addressing concerns about OpenAI's reliance on the New York Times' information, Altman clarified that future AI models will no longer require large datasets from a single source. Instead, these models can be trained on more concise but high-quality data acquired through partnerships with multiple publishers. Altman asserted that their dependence on the New York Times for training data is not a priority and often misunderstood.

Altman also highlighted an emerging trend in AI training, anticipating that future models will be capable of drawing insights from smaller volumes of superior training data. Consequently, it will be possible to gain a comprehensive understanding of a subject without resorting to exhaustive amounts of data. For instance, one doesn't need to study thousands of biology textbooks to grasp high school-level biology.

While focusing on the development of AI technology, Altman acknowledged the need for new economic models that reward those whose expertise and work contribute to training AI models. He suggested that future models could include links to publishers' websites, aligning their interests and promoting a mutually beneficial relationship.