Normal Technology: Powerful AI, but a Tool Nonethless

This article discusses the concept of AI as "normal technology." The author argues that advancements in AI technology do not automatically translate into economic impact or widespread adoption. Instead, adoption tends to lag behind development, as organizations take time to adapt to new models and capabilities. The process of adoption is slowed down by organizational changes, which involve modifications to processes, responsibilities, incentives, and cultural shifts.

In the field of software development, AI has made coding faster, but overall software product development has not sped up due to bottlenecks in decision-making on what should be implemented and its value to the business.

The author also acknowledges that AI adoption is slower in high-risk areas such as healthcare, law, and finance, where mistakes could have severe consequences. Even if AI becomes more reliable, people may remain cautious, particularly when it comes to decision-making roles like AI judges. The key issue is not only reliability but also legitimacy and responsibility.

Additionally, the author suggests that AI can accumulate power indirectly. People already rely on AI to assist with cognitive tasks, and professionals may increasingly depend on AI to analyze situations, propose actions, and recommend decisions. While the final decision may still rest with a human, the AI's influence on that decision becomes harder to discern. This outsourcing of critical thinking could result in AI accumulating power over time.

In terms of job replacement, the author is more pessimistic than previously stated. If an AI system becomes more efficient than a human at performing most tasks associated with a profession, it could fundamentally transform that profession. Humans may shift towards roles where AI is less capable, creating a polarization between deep technical roles and consulting, sales, product management, and AI adoption roles, where human skills remain valuable.

The author ultimately rejects the notion that AI should be treated as a new species. Instead, they view AI as a powerful tool that can be more capable when used by humans, similar to other technologies such as electricity or the internet. However, they emphasize that the largest risks come from the way humans use AI, as powerful tools can be dangerous when utilized by malicious individuals.