Why A.I. Didn't Transform Our Lives in 2025 (www.newyorker.com)

Predictions from OpenAI executives, including CEO Sam Altman, suggested that 2025 would herald the advent of AI agents capable of joining the workforce and performing multi-step tasks autonomously. These agents were envisioned to streamline complex activities such as hotel bookings by navigating various digital platforms and decision-making processes. However, as 2025 approaches, the promised transformation has not materialized. Critics, including OpenAI co-founder Andrej Karpathy, describe existing AI agents as “cognitively lacking,” highlighting significant shortcomings in their ability to perform real-world tasks effectively. The gap between the anticipated capabilities of AI agents and their current performance raises concerns about the future of AI in the workforce. While models like OpenAI's Codex have made strides in automating software development, they struggle with tasks that require maneuvering outside of coding environments, such as utilizing a mouse to interact with web pages. Furthermore, the reliance on large language models (LLMs) for decision-making can lead to compounded errors, which can derail multi-step tasks. As OpenAI shifts its focus back to its core chatbot product, the excitement surrounding AI agents gives way to a more cautious outlook, prompting industry experts to view this era as the "Decade of the Agent" rather than a singular moment of breakthrough.