
Adept is a company focused on building an AI teammate for everyone, which they believe is the future of computing. The San Francisco-based company is building a machine learning model that can interact with everything on your computer, and they're introducing their first large model, Action Transformer (ACT-1), to achieve this goal.
ACT-1 is a large-scale transformer trained to use digital tools and is designed and trained specifically for taking actions on computers in response to natural language commands. It is the first step towards a foundation model that can use every software tool, API, and website that exists. With ACT-1, Adept aims to create an entirely new way to get things done, taking users' goals, in plain language, and turning them into actions on the software they use every day.
The company's vision is to create an AI teammate for everyone, enabling more informed decisions, discovering new solutions, and giving more time for the work people love. They believe that AI systems should be built with users at the center, where machines work together with people in the driver's seat.
ACT-1 can take a high-level user request and execute it, using a custom "rendering" of the browser viewport that's meant to generalize across websites. The action space is the UI elements available on the page, and ACT-1 can observe what's happening in the browser and take certain actions like clicking, typing, and scrolling. In addition to these capabilities, Adept expects to improve ACT-1's speed on the modeling and software side, making it faster with latency that's largely imperceptible to humans.
Adept's target market is individuals and businesses looking for a new and innovative way to interact with their computers. The company's focus on natural language interfaces that allow users to tell their computers what they want directly is a potential game-changer in the computing industry.
In conclusion, Adept is a San Francisco-based company building an AI teammate for everyone, focused on creating an entirely new way to get things done. They're introducing their first large model, ACT-1, which is designed to take actions on computers in response to natural language commands, and is the first step towards creating a foundation model that can use every software tool, API, and website that exists. Adept's vision is to create AI systems built with users at the center, where machines work together with people to discover new solutions, make more informed decisions, and give more time for the work they love.