OpenAI announces ChatGPT agent for web browsing
Meet the new OpenAi web browsing agent.
On Thursday, Openai announced the Chatgpt agent, a tool capable of navigating the web and performing tasks on your behalf. Taked in a post X before Livestream, the Chatgpt agent combines the autonomous capacities of his operator agent and the intelligence of reasoning of his deep research tool.
The operator of Openai, which was launched in January in preview mode for Chatgpt Pro users, could scroll, click and type on the web, but has had limits and has never seen a generalized version. In-depth research is another type of agent who can search for the web and compile information in the name of the user, but he could not take measures beyond. The launch of the new OpenAi web browsing agent effectively combines the two tools.
Credit: OPENAI
I test the new comet of a navigator Ai de Perplexity. Here are 3 features that I tried right away.
“The operator could not dive in depth in the analysis or write detailed reports, and in -depth research could not interact with websites to refine the results or access the content requiring user authentication,” said the OPENAI ad. “We have seen that many questions that the users attempted with the operator were in fact better suited to in -depth research, so we addressed the best of both.”
The new Openai tool is part of the world of rapidly growing agentics
OPENAI’s Chatgpt agent joins other recently published agent tools which can perform tasks in the name of the user. Although it is not a full web browser, it acts in a similar way to the Assistant de brower by Perplexity Comet. Anthropic also has a tool called “computer use” which can take control of your cursor and write code. As the models become more advanced, they are more able to perform autonomous tasks. Navigation on the web is considered one of the next arenas for AI laboratories to compete, with Openai, Anthropic, Perplexity already shipped the features and the Google Mariner Project research prototype.
Mashable lighting speed
In agent mode, you can ask Chatgpt to do tasks such as “look at my calendar and put me on the meetings of customers to come according to recent news” or “plan and buy ingredients to prepare a Japanese breakfast for four”, according to an Openai spokesperson.
OPENAI said that users are still in control and that Chatgpt requests authorization before taking measures, such as forming for forms, purchasing or managing personal information. OPENAI said users can easily interrupt or take over if necessary. Certain tasks, such as sending emails, require surveillance called watch mode.
On the security front, the chatgpt agent is not authorized to perform “high -risk” tasks such as financial transactions or legal advice, and it is trained to recognize malware or opposing attacks of rapid injections or other manipulative tactics.
In terms of data collection, Openai said that Chatgpt can delete their navigation data and disconnect websites with one click and not collect data while the user is actively involved in tasks such as password entry.
Given progress, there is also a considerably higher risk than things are wrong. OPENAI classified the Chatgpt agent as a high risk in its preparation for biological and chemical capacities.
There are good news for users who do not have the $ 200 plan plan for Chatgpt Pro. It is available for PRO users, more and the team. PRO users have access today and more users and the team have access to the next few days. Professional users receive 400 messages per month, while the more, the team and other paid users receive 40 messages per month.
Disclosure: Ziff Davis, Mashable’s parent company, in April, filed a complaint against Openai, alleging that it has violated Ziff Davis Copyrights in the training and exploitation of its AI systems.
