While the major technological developers take place to deploy the next generation of artificial intelligence, some people already have an army of AI assistants to check their daily task lists.
These are called agents: AI systems which automatically offer a plan and execute it at your request. They can take the form of a sales assistant, a video game coach or even a fully independent car.
“This notion that we will have our own personalized assistant, if you want, is a way of thinking about it,” said Tammy Madsen, professor of strategy and innovation at the University of Santa Clara, about the rise of These specialized virtual workers.
Last month, Openai, the company behind Chatgpt, presented an AI agent called operator which she announced as capable of using the web to perform tasks such as filling the forms and ordering grocery products for a certain recipe. In a demo From the agent, Openai showed an operator reservation for dinner, buying tickets for a basketball game and commander of pizzas, each time the specific preferences of the user.
Google has also announced the launch of a AI co-scientificthat the company has described as a “virtual scientific collaborator” made up of several AI agents.
Companies are also looking forward to adopting them, using AI -powered tools to automate tasks traditionally managed by customer support representatives, job recruiters, travel agents and others.
During a showcase last year, Salesforce introduced Agentforce, a platform that allows companies to create their own AI agents. Shortly after, its Rival Microsoft announced its own plan to allow companies to do the same with Copilot Studio, the complete platform of the technology giant for the construction of AI agents.
Many in technological circles have already documented technology growth capacities. Aaron Levie, CEO of Cloud Company Company, published a video on X last month showing how the operator of Openai could take the initial measures necessary to exchange sensitive documents with another company.
AI agents are a step beyond the capacities of AI chatbots, which fueled the generative AI boom which started with the launch of the model of the large language model of Openai in November 2022. LLM as the chatgpt are predictive text models, which means that they generate answers to human entry, but not I do not take independent measures.
In recent years, Openai and Tech Titans like Google and Meta, as well as the main IA startups like Anthropic and Stability IA, have run to develop multimodal LLM capable of generating an outing in visual supports such as images and video.
But the developers today are also increasingly integrated LLM into a growing border of tools fueled by AI: agents who can make decisions and take autonomous measures on behalf of their users.
The booming accessibility of this technology means that people can use various agents to automatically transcribe and summarize work meetings, reserve a flight, register for a exercise course, serve as a health coach or even manage their share portfolio.
“A big question, as individuals as an individual, we use more and more these tools, is: do we lose critical thinking capacities over time?” Said Madsen. “We can automate certain routine tasks, but is this additional capacity in our brain now used in another way in order to continue to advance our knowledge, our critical thinking? Or do we become relaxed and dependent on the tools?
AI assistants and agents have even slipped into the advertisements of the starred super bowl of this year, as when Salesforce had Matthew McConaughey Praise the capacity of his agent AI To reserve restaurant reservations.
“The big impact on the company has positive and negative aspects,” said Param Singh, professor of commercial technology and marketing at Carnegie Mellon University. “If you automate a large part of these tasks, you will be more effective as a business. But these tend to be the tasks of entry-level jobs, and these types of jobs will be taken up by the much more AI. »»
But he said that in this stage, human workers will always be necessary to supervise and guide automated agents, who cannot be invoked to perfectly interpret the instructions or to understand the necessary context by themselves.
These tools can also be subject to misadventures – as when a driver -free Waymo Robotaxi has led his client to circles instead of taking him to the airport – which does not occur under the control of human workers.
Singh said that AI agents work reliably for repetitive tasks carried out in a controlled environment, where they are only allowed to make a small set of decisions and that any error would not lead to consequences Critics.
“If it is a decision with high issues, then we must be very cautious, because even an error may have a significant impact,” he said. “If these [AI agents] are in any type of environment, even slightly dynamic, so I don’t think we can trust them at the moment. »»
Despite their popularity in hot airlines in recent years, AI agents have been changing for some time. Vocal assistants like Siri and Alexa from Amazon, as well as smart home apparatuses like Google Home, were among the first AI Assistants to have been adopted by the masses. But even these devices, which could process vocal orders and perform tasks, lacked all the autonomy and decision -making capacities of today’s AI agents.
While these tools become more sophisticated and further infiltrated daily routines, some experts say that people are likely to become too dependent on the aid fueled by AI to allow them to spend life. Others are more optimistic, saying that the adoption of AI agents for simple tasks could make critical thinking and creativity all the more important.
“Recognizing that it is a collaboration between you and the agent is a good starting point,” said Madsen, professor of the University of Santa Clara. “You are the one who should guide the agent to some extent, as opposed to the agent who guides you.”
