AI Progress Stalling?

Recent reports suggest that AI development may be slowing down, with OpenAI's GPT-5 model showing less progress than anticipated. Some experts believe that the current deep learning techniques are reaching their limits, and there may not be enough fresh data to continue scaling AI systems effectively. This challenges the assumption that simply adding more computing power and data guarantees better models.

Despite improvements in areas like coding, writing, and reducing factual errors, GPT-5's performance has led to concerns about diminishing returns on investment in AI. The model employs a system of specialized components and a real-time router to optimise speed and efficiency, but the initial launch faced challenges, including a crash of the automatic switcher. While GPT-5 demonstrates competence and improvements in instruction following and tool use, some argue that its practical usefulness over previous models is not significant enough to justify the continued exponential spending in the field.

Industry leaders suggest that future progress will require deeper breakthroughs rather than just scaling existing models. This potential plateau in AI advancement raises questions about the long-term viability of the current development approach and its impact on related sectors.