Toni Cowan-Brown and Benedict Evans discuss the current state and future of AI, particularly focusing on Large Language Models (LLMs) like ChatGPT. Evans introduces the concept of "unbundling AI," suggesting that while LLMs appear to be a generalized computing platform, their broad applicability may lead to a need for more specialized tools and interfaces. They explore the limitations of LLMs, such as error rates and the challenge of using a question-answer model for tasks requiring direct control and editing. The conversation touches on the paradox of general-purpose UIs versus specific task requirements, the evolution of tools from simple to complex, and the potential for LLMs to become integrated into vertical solutions and everyday software features, often invisibly to the user. They also consider whether the current conversational interface is the most effective way to interact with AI, drawing parallels with the limited success of pen computing.
Sign in to continue reading, translating and more.
Continue