AI companies are pivoting from creating gods to building products. Good.
AI Snake Oil Arvind Narayanan
AI companies initially pursued two flawed approaches to commercialization—OpenAI and Anthropic focused on research and models while Google and Microsoft hastily integrated AI into existing products—but are now shifting toward building proper consumer products. The cost per unit of capability has dropped by a factor of over 100 in the last 18 months, yet companies still face five major barriers to commercial success: cost, reliability, privacy, security, and user interface design. Developers must recognize that consumers expect AI to behave like deterministic software rather than adapt to AI's limitations, requiring solutions to technical and sociotechnical challenges that will likely take a decade or more to resolve.
Why it matters
Turning models into products runs into five challenges
