Google’s latest announcement of Gemini 2.0 brought with it the unveiling of several AI Agents in progress, the most exciting of which may be Project Mariner.
This AI agent claims to be able to control your computer without the need for extensive human input.
In this article we’ll tell you all about what Google's Project Mariner is, how to use it and if it's safe.
What is Mariner?
Project Mariner is the name of Google’s new AI agent that can control your computer. It functions as an extension within the chrome browser .
Mariner can automatically navigate websites, click on links and fill out forms. This means it can perform web based tasks like a human, such as comparing prices or researching information.
Google says that Mariner can ‘understand and reason across’ any information from within the browser screen, this includes ‘ pixels and web elements like text, code, images and forms.’ It also displays a clear view of its plan and following actions, allowing the user to understand the decision-making process.
Project Mariner has been evaluated against the WebVoyager benchmark. This isa dataset of tasks that are designed to evaluate the capabilities of AI agents in interacting with real websites. Mariner in its current form achieved 83.5% against the WebVoyager benchmark.
How to use Mariner?
Project Mariner is currently an early research prototype. This means that it isn’t publicly available yet as Google is still actively refining the project. However, if you can’t wait to try out Mariner you can apply to be one of Google’s ‘Trusted Testers’.
1 .Visit the Project Mariner website and click ‘Join the trusted tester waitlist.’
2. Fill out the Google form with your details.
3. Ensure that you read the Google Terms of Service and Privacy Policy.
4. Click submit.
5. Await Google’s response.
Is Mariner safe?
Allowing an AI to manually use your computer is an intimidating concept.
The idea that an AI is able to perform tasks like a human is exciting and may allow mundane digital chores to be automated and more complicated projects to be made easier.
However, as this new technology can interact with anything on your desktop there are understandable concerns about security, privacy, and the potential for unintended consequences.
To mitigate the risk Google must take significant steps to ensure the security of users data and be transparent with the safety protocols in place.
The tech giant states that in the development of these new technologies they ‘recognize the responsibility it entails, and aim to prioritize safety and security in all our efforts.’
Though they don't confirm the specifics, Google say that to ensure safe and responsible development they are taking an ‘exploratory and gradual approach’. This means conducting research on multiple prototypes, as well as iteratively implementing safety training. Before public release they are also working with trusted testers and external experts as well as performing an extensive risk assessment.
