A few weeks ago, Flo Crivello, the CEO of a startup called Lindy, asked his personal assistant Lindy to change the duration of an upcoming meeting from 30 to 45 minutes. Lindy, an AI-powered software agent developed by Crivello’s startup, extended all the 30-minute meetings on his calendar, causing havoc.
Crivello’s company, along with other startups, aims to develop AI agents that can perform useful tasks beyond generating text like chatbots do. The goal is to create AI agents that can help people with everyday chores. For example, in addition to providing planning advice for a business trip, these agents might be able to find and book suitable flights, make the payment using a company credit card, and fill out the necessary expense report.
However, there is a catch. As Crivello’s calendar mishap demonstrates, these agents can sometimes become confused and make embarrassing or costly mistakes. Nobody wants a personal assistant that books a flight with multiple layovers just to save a few dollars, or schedules conflicting appointments.
Although Lindy is currently in private beta and Crivello claims that the calendar issue has been resolved, there is no fixed timeline for the product release. Nevertheless, he believes that AI agents like Lindy will become commonplace in the near future. Crivello optimistically predicts that within the next two to three years, these models will become significantly more advanced.
The concept of AI helpers that can take actions on behalf of users is not new. Popular examples include Apple’s Siri and Amazon’s Alexa. However, the recent release of ChatGPT has given rise to the idea that it might now be possible to create AI agents with broader capabilities. Some programmers and entrepreneurs have started their own projects using large language models, such as the one behind ChatGPT, to build more advanced AI agents.
After seeing discussions about ChatGPT’s potential on Twitter, programmer Silen Naihin joined an open-source project called Auto-GPT, which provides programming tools for building AI agents. Naihin has previous experience in robotic process automation, a simpler form of automating repetitive tasks on a computer widely used in the IT industry.
