Artificial intelligence runs fast, offering new opportunities to key sectors such as business, entertainment and health. In particular, the use of AI can prove to be basic in facing some of the main challenges in the sector, such as the lack of health personnel. In this sense, Daniele Davoli, CEO of Senseledge – startups who brings together health professionals and artificial intelligence experts – explains: “The lack of operators represents an increasing challenge for health systems all over the world, and Italy is no exception. Integrating the use of AI, which can accelerate and improve the provision of health services, contributes to making the entire national health system more efficient”.
Continuous and frenetic innovation, therefore. Well beyond Openai chatgpt. These peculiarities that embrace the words of CEO of Nvidia, Jensen Huang, who in Las Vegas – protagonist of the highlight at the Consumer Electronic Show (the main fair in the world dedicated to consumer electronics), spoke of the three laws of scalability. Premising that “the more data you have, the greater the model” and “the more calculations they apply, the more the model becomes effective”, Huang has dealt with the theme of the so -called “third law”: the one that is allowing us to enter the Agentic AI era.
Let’s talk about a new concept of artificial intelligence. Better still, of a software designed to interact with data and tools so as to provide for the minimum intervention by the user. But is it concretely a revolution or is it, more simply, a new label for existing concepts? Today.it asked Dario Maggioreini, director of Techbusiness and associate professor of the University of Milan. Which replies: “Scientific literature has been talking about intelligent agents for decades, and this is perhaps the historical moment in which we can get them. Certainly there is a marketing aspect to keep the attention of investors high; Nonetheless, the risk is that more time may be necessary than the market is willing to wait “.
Artificial intelligences compared
It is easy to say classical, generative or agent artificial intelligence. With Maggioreini (who spoke, in unsuspected times, with Today.it, of the risk of dossier in Italy) we take stock of what are the various definitions. “The” classic “artificial intelligence aims to solve a specific problem in which the answer is typically bound to a choice (something like:” Is this the photo of a dog? “Or” what side is it better to go? “)”.
Instead – the expert resumes – “With the generative, the idea is to create content by drawing information from a series of existing sources. A lot of images and video has been talked about, but for a long time it has been done with the text and music. The real revolution is the fact that it is fueled by a prompt; that is, interpreting a phrase of totally arbitrary description”.
And what about the AGENTIC AI (which Gartner has included among the ten Tech trends of 2025)? “In this case – Majorini replies – it aims to a much more complex result. The goal is to create an entity, an agent, who confronts the world around him, adapts to you by continuing to learn and is able to perform actions to obtain a result. Unlike previous systems, an agent becomes progressively more skilled in operating and independently decides his objectives based on the context. interacts with us and makes us proposals to satisfy our needs “.
Technological implications
Analysts provide that, by 2028, Ai agents will take on at least 15 percent of work decisions in companies. In the practical act, the director of Techbusiness resumes “an agent can automate complex operations in which today’s to today are still in difficulty. For example, he can move in critical environments – such as those with toxic fumes or radiation -, overcoming complex obstacles without putting a human being at risk. There are also interesting potential, in the medical field, thanks to their interoperability with other systems and in the services of Help Desk thanks to an interface. which can be almost natural “.
Nonetheless, it seems that the Agentic Ai has not yet been made. “It is just like that – admits Maggioreini -, we know that today, finally, we have a sufficient calculation capacity (the famous GPUs, units for parallel processing designed above all in the graphic area) to do it. In any case, to build an agent it is necessary to combine and coordinate different artificial intelligence techniques, both modern – such as the LLM (the Large Language Model capable of recognizing and generating text) – unfortunately classic. However, at present, it is not yet completely clear how to obtain this coordination nor if new AI techniques must be developed to be accompanied by the existing ones “.