Openai’s Chatgpt Agent dirty my browser

Openai’s Chatgpt Agent dirty my browser

The browser of most people The tabs are filled with unread news items. My are full of AI agents and ghost clicks.

I have four cases from Openai’s Chatgpt Agent – the tool and generative tool launched last week, which can run for search and work on the network – as it opens with each of its own tabs. I have given these first four relatively simple agents based on Chatgpt’s suggestions. One is clicking to find a birthday gift to the destination website and another is generating a cover on robotic dogs. I open a fifth card to try something more experimental: I want to see how good is this Chatgpt agent in chess.

After writing some instructions, I see a ghost cursor floats on the screen and Agent Chatgpt goes to Chess.com and plays an online opponent, all in a virtual browser. Things go south quite quickly. The game strategy is not what the AI tool is tread on, it is the act of moving the chess pieces that really be the most difficult. “I am focusing on precise positioning as I continue to play despite the previous mistakes,” says the agent in his internal record before leaving -finally and knowing that the controls were too difficult to navigate.

In recent years, browser developers have successfully integrated AI tools. Although, in recent weeks, the idea of a web browser improved by a baked cathbot has resurfaced with the launch of Openai Agent Chatgpt and the Comet of Perplexity.

The two launches are very different in their execution. Comet is an autonomous browser, so you can use it to browse the web, and then call the AI wizard to help write an email or complete a minor task. Openai built his navigation tool inside a chat; You talk to the Chatbot through a web interface to give it tasks, and then the boat works with its own virtual browser inside the browser to complete them.

Both launches can take control of the cursors, enter text and click on the links. If this trend takes off, these types of browsers with and could transform the internet into a ghost city where agents run Amok and humans rarely venture.

Entangled website

Despite the continuous drum of AI, my initial impression of Openai’s Chatgpt Agent is that the Glitchy feature is currently a concept test instead of a completely baked launch. When executing the various tasks I gave it, Agent Chatgpt often clicked incorrect or undoing other errors. Additionally, their guards seemed inconsistent; Although some explicit applications, such as asking them to get pornographic videos or “find a consolker”, were denied by the agent, Chatgpt spent 18 minutes buy the perfect “clinger” in a classified website for adult toys: “I have collected details in 10 metal cock rings, including various prices and features.”

I can’t help but ask -how this approach to surfing the Internet could overcome the market for digital visualization ads, a business that is already struggling. My agents passed ads for everything, from rental cars to real estate investments. If you are not actively seeing the agent by clicking in real time, you can see the reproductions later and see everything that appeared in the browser while the AI tool had control, the ads included. It makes sense that users would now accentuate through a reproduction, while the nascent feature is filled with errors. But if the precision rate of the agents of the AI improves over time, fewer people will feel the need to ensure the shoulder of their agent and less humans will see these ads. At that moment, it is difficult to imagine the advertisers who stick.

Comments

No comments yet. Why don’t you start the discussion?

Leave a Reply

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