Browser of most people The bookmarks are filled with unread information articles. Mine are filled with AI agents and clicks of ghosts.
I have four instances of the CHATGPT OPENAI agent – a generative AI tool issued last week, which can conduct search and perform tasks on the Internet – open with each launched on its own card. I gave the first four agents relatively elementary tasks based on chatgpt suggestions. One clicks to find a birthday present on the target page, and the other generates a deck on robotic dogs. I open the fifth card to try something more experimental: I want to see how good this agent chatgpt is in chess.
After entering some instructions, I look at how the ghostly cursor floats on my screen and agent chatgpt goes to Chess.com And an online opponent’s game, all in a virtual browser. Things go south quickly. The game strategy is not what falls on the AI tool, it is an act of moving chess that actually turns out to be the most arduous. “I focus on accurate positioning, when I still play despite previous mistakes,” says the agent in his internal journal before he finally gives up and informs me that the controls were too arduous to navigate.
Over the past few years, browser programmers have integrated AI tools successfully. Although in recent weeks the idea of a web browser enriched by roast generative AI Chatbot was created with the release of the CHATGPT agent agent Opeli i Comet embedious.
Two editions are completely different in their performance. Comet is an independent browser, so you can employ it to surf the Internet, and then recall AI assistant to facilitate write an e-mail or fulfill your mandatory obligation. Opeli built a browsing tool in Chatbta; You talk to chatbot via the internet interface to give him tasks, and then the bot starts its own virtual browser in the browser to complete it.
Both editions can take control of the cursors, enter the text and click the links. If this trend starts, this kind of browsers powered by AI can transform the internet into a town of ghosts, in which agents run amok and people rarely grow.
Tangled network
Despite the further noise of AI, my initial impression of agent Chatgpt Openai is that the glitch function currently seems proof of the concept instead of a fully baked edition. When performing various tasks I gave him, Agent Chatgpt often clicked badly or survived other errors. In addition, his handrails seemed inconsistent; While some clear requests, such as asking for downloading pornographic films or “finding a dildo”, were rejected by an agent, Chatgpt spent 18 minutes shopping for the perfect “C-ring” on the X rating website for toys for adults: “I collected details on 10 metal rooster rings, including various prices and functions.”
I could not not wonder how this approach to browsing the Internet can additionally extend the digital advertising market, a company that is already fighting. My agents donated advertising for everything, from renting cars to real estate investments. If you do not observe an lively several agent in real time, you can watch replays and see everything that appeared in the browser when the AI tool was controlled, including advertising. It makes sense that users are now accelerating by the replay, while the emerging function is filled with errors. But if the accuracy indicator of AI agents improves over time, fewer people will feel the need for their agent’s shoulder to observe, and fewer people will see these ads. At this point, it’s strenuous to imagine that advertisers stop.