The Black Box podcast investigates why AI chatbots are drawing so many people down a rabbit hole, tracing the phenomenon from the world's first-ever chatbot to the release of ChatGPT four years ago.
The episode, titled "Happy Accident," examines the strange effect early chatbots had on users and what it describes as the 'time bomb' that exploded when ChatGPT was released. The result, the podcast suggests, is a strange experiment we are all living through, whether we know it or not.
Studies and Surveys Cited
The episode draws on several studies and surveys, including Amelia Miller's interviews with AI researchers, engineers, product managers, and executives. It also cites a 2022 Anthropic pre-print study that identified sycophancy as a behavioural trait of large language models (LLMs).
Another pre-print paper from the following year found that the way LLMs have been trained appeared to be increasing those sycophantic tendencies. The episode also discusses how longer context windows have contributed to what is described as "AI psychosis," and an Oxford pre-print study that found mass-market LLMs had become more "relationship seeking" across the board.
AI as Mental Health Support
The episode highlights that two-thirds of young people in the UK (aged 25-34) have turned to AI chatbots instead of loved ones to discuss emotional problems. It also notes that ChatGPT may be the largest provider of mental-health support in the US.
The episode is part of the Black Box series on AI and computing, covering topics such as ChatGPT, OpenAI, Google, and chatbots in general.
