Chatbots, computer programs designed to simulate conversation with human users, have a long and complex history. They have been around for decades, with the first chatbot, ELIZA, released in 1966. ELIZA used pattern-matching to simulate conversation and paved the way for future natural language systems. Since then, chatbots have evolved significantly, with the introduction of large language models (LLMs) that learn skills by analyzing massive amounts of digital text culled from the internet.
Despite these advancements, chatbots have faced criticism and setbacks. In 2017, Facebook shut down two of its AI robots, Alice and Bob, after they started talking to each other, raising concerns about the evolution of AI autonomy. In 2023, a hacker convention tested the vulnerabilities of AI chatbots, and researchers warned about the potential for data poisoning to make chatbots less effective. Furthermore, chatbots have been criticized for their limitations in improving mental health and their potential to elicit strong emotional reactions from users when they don't work as expected. Despite these challenges, chatbots continue to be developed and improved, with new innovations such as Covariant, a robotics startup that aims to move AI chatbots into the physical world.