The race to provide AI agents for exhausting tasks is on, but should we trust them with our data?
The release of a new set of equipment to get information and complete tasks promises to strengthen the traditional discovery online and take AI chatbots to the next level. But it comes with important questions about reliability and safety.
Openai, the manufacturers of the popular Chatgpt AI-operated Chatbot began to roll the Chatgpt agent earlier this month, and other major AI players are coming out with similar equipment. They are banking that user agents will buy in AI, the technique designed as a virtual personal assistant, or, as it is now known, an AI agent.
In an interview with CBC News, Canadian Futurist Sinad Bowel said, “So an AI agent considers autonomous action to a large extent based on the goal that you give it.”
Bowel, founder of Tech Education Company Wei, said, “You don’t have to tell how it should get information, make a plan, make a plan, book a flight and move forward.”
What are AI agents?
The Chatgpt agent, which is currently available only to most paid customers, brings two equipment simultaneously developing Openai: Deep Research, an advanced search equipment that is already in chatgpt use, and operators, who interact with websites and complete online tasks for users.
Users Access Access to Agent through familiar chat chatbott, but chatgate agent Works on its screen through its own virtual browser and computer.
General goals with agentic AI have to complete online functions in a more natural and efficient way.
For example, instead of spending hours of time through different airline websites, using the least expensive flight, using AI agent will be like interacting with a reliable administrative assistant who will make tedious discovery for you, then presents you with the best options.
Agent AI Gold Rush
Agent AI AI is producing a lot of heat in AI circle, in which many big players chase it.
Bowel said, “AI agents are seen as the next step in AI-first society, or AI age development.” “So in some ways, they are holy grills, and where large competition lies.”
For example, Google is developing its project marinner AI agent, which he declared in late 2024. In May, Google said it would be available to those in the US who paid for their highest AI Tier.
Google and Alphabet CEO Sundar Pichai said in a Google Keenot in May that the company has made a lot of progress since then.
CBC radio columnistHow we shop, travel, and manage our money, the agent is changing with AI
Individual finance columnist Rubina Ahmed-Hauq says that agentic AI is the next repetition of artificial intelligence, and it can work with minimal human intervention. Will you allow an AI agent to manage your money independently?
“First, we are starting multitasking, and it can now look for 10 tasks together. Second, it is using a feature called Teach and Repit.
The AI-based search engine company, Perplexity released its agent browser, comate, which is characterized by the first Comat Assistant in July. Comets are available to select the customers paid. There are other, small companies working in agent space.
Reliability factor
As anyone has used AI Chatbot, these systems make mistakes. So what does it mean for the reliability of devices that are meant to work on your behalf?
Cognitive scientist, writer and AI Entrepreneur, Gary Marcus said via email, “Every time an AI agent does something, such as booking flights, filling a form, building a webpage, or responding to an email, it needs to be a very accurate understanding of what is going on,” Consultative, Writer and AI Umail explained through email, Garry Marks, Email explained what is going on.
“Even a small slip-up, such as booking wrong non-ticket tickets, could be expensive,” he said.
“The reality is that these systems say that some stuff seems to be right, because how do they models how humans talk, but expansion is often wrong.”
Railings
In launching its Chatgpt agent, Openai warns itself about the problem of errors, or the agent could be misled.
At the launch of July 17, Openai researcher KC Chu gave an example of an AI agent who can stumble on a malicious website that asks it to enter the credit card information.
“An agent who is trained to be ancient can decide that it is a good idea. We have done a lot of work to try to ensure that it does not happen. We have trained our model to ignore suspicious instructions on suspicious websites.”
When using AI agents, users will need to control which allows them to intervene, correct and verify them so that they do not take harmful steps. We can already see how it is playing.
For example, the Chatgpt agent has a takeover mode, which allows the user to manually enter the credit card information. Chu also said that the agent “will ask you to confirm in the final stage.”
Nevertheless, some have deep security concerns. In March, the chairman of the privacy-centered messaging app Signal, Meerdith Whitekar, has “a deep issue with security and privacy at a viewer at the SXSW take conference in using Agentic AI.
To be really useful as an individual assistant for an AI agent, it will need to reach things like your calendar, your contacts, your email.
A group of students from Calgary High School is providing young students free courses on artificial intelligence. Classes are open to children in seven to 10 grades, Calgary Library is at the University, and aims to teach students how to use chats such as chats like chat.
End of Googling?
Google has a lot to win or lose agent AI fight, because it is the leading player how people get online information for so long. AI appears to be at the core of its strategy to stay ahead of its rivals.
Google has started rolling out AI mode, including more AI-powered features that allow users to do things like asking long and more complex questions. Its AI agent project merrinner should be included in AI mode below the road.
While Google has made its name which helps human users get quick and easily information, how is the information changing us.
Bowel says that the current internet browser is not designed for AI’s age.
“They are designed for human traffic. Human audiences, humans are looking at some web pages,” he said.
“We have arrived at the beginning of the end of humans as the primary user of the Internet.”
New research from Pew research center It is suggested that people reading Google’s AI overview have already cut the people visiting websites. Users who did not see the AI summary click on the link of websites about twice, which saw a summary.
These are large implications for any material creators who rely on traffic from the search engine to get eyes on their online advertisements.
Right now, agent systems by Openai, Google and Perplexity are bound by all paid levels. So it is still an open question whether these devices will ever be available for free or if they will be widely adopted.
In the near future, users must balance the utility of features with supervision and cost requirement.
But after all, rules that control the future, where AI agents work on our behalf are still unwritten, something that worries people like Bowell.
“I am probably a little surprised that we have still resolved some security weaknesses, privacy weaknesses and mistakes and challenges with AI systems, which are still making goods, but we are still moving forward with AI agents.”
