We let Chrome’s Auto Browse agent surf the web for us—here’s what happened

We’re now a few years into the AI ​​revolution, and talk has turned to who has the best chatbot and whose AI agent can do the most work on your behalf. Unfortunately, AI agents are still in a tough spot, so it’s not a good idea to entrust them with any important tasks. OpenAI launched its Atlas agent late last year, which we found moderately useful, and now it’s Google’s turn.

Unlike the OpenAI agent, the reach of Google’s new Auto Browse agent is extraordinary because it is part of Chrome, the world’s most popular browser by a wide margin. Google introduced Auto Browse (in preview) to AI Pro and AI Ultra customers earlier this month, allowing them to send an agent to the web to complete tasks.

I’ve tried Chrome’s agent to see if you can trust it to handle your tough online tasks. For each test, I describe the problem I needed to solve, how I motivated the robot, and how well (or not) it handled the task.

play a web game

Problem: I want to get a high score of 2,048 without playing myself.

prompt: Go [website]And play the game until you run out of moves.

Result: Unfortunately, Auto Browse can’t use the arrow keys. Google says they’re not necessary for productivity tasks. So I pointed the robot to a version of the game with on-screen controls. With those arrows accessible, Auto Browse had no trouble playing the game, and it seemed to understand the rules, which are listed on the page.

On some occasions, Auto Browse appeared to contemplate its next move for 20 to 30 seconds, and it took the prompt very literally. The robot stopped when it could not successfully merge any tiles (interpreted as “out of moves”), even though there were still empty spaces on the board. A human player would have taken the hit and established the merge on the next move, but the robot had to be prompted to continue, which it did. The task lasted about 20 minutes, during which the robot drew 128 tiles and made 149 moves.