Adept, an artificial intelligence (AI) company in California, has unveiled the Action Transformer (ACT-1). When you instruct it with a written or verbal command, this AI model can use web browsers like a regular human would to accomplish its task.

Adept shared a video on Twitter to demo how ACT-1 works. In it, the user submitted a written command “Find me a house in Houston that works for a family of 4. My budget is 600K ”on a text box, after which ACT-1 promptly searched for homes on a real estate website using the specified criteria.

Other demonstrations also showed ACT-1 creating records on a customer relationship management platform, adding new data and records on a spreadsheet, finding items for sale across multiple tools/websites, and looking up information on Wikipedia.


The reason why ACT-1 is capable of scrolling, clicking, typing, and doing other interactions with digital tools is that it’s a transformer model. Basically, it’s a neural network that learns from existing data and examples and tracks relationships among elements in the data set.

In its introductory blog post for ACT-1, Adept believes that interactions in the future will shift towards natural language interfaces instead of the graphical user interface that’s common today. The company also recognizes potential misuse of AI, which is why it plans to use “careful, staged deployment” along with machine learning techniques.

Adept has a waitlist page at for users who’d like to gain access the alpha release of the company as they become available.

Join the Conversation

1 Comment

Your email address will not be published. Required fields are marked *