{"id":5093,"date":"2026-10-07T11:58:14","date_gmt":"2026-10-07T09:58:14","guid":{"rendered":"https:\/\/fondazione-fair.it\/?p=5093"},"modified":"2026-10-07T12:06:05","modified_gmt":"2026-10-07T10:06:05","slug":"ai-in-italy-bringing-research-into-business-and-society","status":"publish","type":"post","link":"https:\/\/fondazione-fair.it\/en\/news\/ai-in-italy-bringing-research-into-business-and-society\/","title":{"rendered":"AI in Italy: Bringing Research into Business and Society"},"content":{"rendered":"

An interview with Daniele Nardi, Full Professor of Artificial Intelligence at Sapienza University of Rome and Board Member of the FAIR Foundation<\/em><\/p>\n

\"Prof.
Prof. Daniele Nardi<\/figcaption><\/figure>\n

Turning Italy\u2019s expertise in artificial intelligence research into applications for businesses and public administration requires training, investment, infrastructure and collaboration between researchers and industry. This transition is the focus of our conversation with Daniele Nardi, <\/strong>
\nFull Professor of Artificial Intelligence at Sapienza University of Rome and Board Member of the FAIR Foundation.<\/strong>
\nOn 5 October, Nardi represented FAIR at \u201cBusinesses in the AI Era: Navigating the New Digital Ecosystem<\/strong>\u201d, held at the Sala dei Notari in Perugia, contributing to the session on AI policies and strategies. The event opened AIxIA 2026<\/strong>, the 24th International Conference of the Italian Association for Artificial Intelligence, taking place from 6 to 9 October<\/strong>. Building on that discussion, we explore what is needed to translate research into practical applications and how FAIR can contribute.<\/p>\n

 <\/p>\n

What are the main barriers today to translating AI research findings into practical use in businesses and public administration?<\/strong><\/p>\n

\u201cFirst of all, I would say the issue reflects a broader wait-and-see attitude among Italian companies, which tend to be reluctant to embrace innovation,\u201d Professor Nardi explains. \u201cThis tendency is also evident in public administration and points to a relatively conservative approach across the country.
\n\u201cThen there are the challenges faced by startups in particular, which are currently among the organisations best placed to develop innovative solutions. In Italy, however, they encounter well-known barriers, especially in accessing capital, and face a slow and complex bootstrapping phase.\u201d<\/p>\n

When an AI system operates in the real world, what assessments are needed to evaluate its reliability and determine how much we can entrust to it?<\/strong><\/p>\n

\u201cFirst, AI is not limited to generative AI. For various more conventional AI applications, there are evaluation methods based on different types of error statistics,\u201d Nardi continues. \u201cThese methods rely on the system being deterministic: it always produces the same output for the same input.<\/p>\n

\u201cWith generative AI systems in particular, however, this property no longer holds. The same input can produce different responses, ranging from slightly different to substantially different. Alternatively, small changes to the input that preserve the meaning of the request can still lead to different responses.<\/p>\n

\u201cConsider a system that processes text: small variations can produce different results, and even the same text can lead to different outputs. This clearly makes these systems very difficult to evaluate. Statistical methods would need to be extended to account for the vast number of possible variations in text, which are typically not standardised. It is therefore difficult to establish a sound basis for statistical assessment.\u201d<\/p>\n

Partly because we do not know exactly what is happening inside the system.<\/strong><\/p>\n

\u201cYes. These systems, particularly generative AI systems, though not exclusively, cannot provide fully reliable explanations of the mechanism that led to a particular response. This makes evaluation even more complex.
\n\u201cThis applies to generative AI, but also more broadly to neural networks,\u201d Nardi explains, \u201cespecially deep neural networks, because the methods developed so far to trace and understand the process behind a particular decision are not yet sufficiently mature.\u201d<\/p>\n

How does this change our own role? What responsibilities remain with us?<\/strong><\/p>\n

\u201cWhen it comes to generative AI, the distinction I made earlier captures, in a sense, the different approach we need to take when considering these systems\u2019 responses,\u201d Nardi continues.
\n\u201cWith a deterministic system, such as a booking system, if a seat is available, the computer tells me it is available: the answer provided by a deterministic algorithm is unambiguous.
\n\u201cThe same does not apply to a generative AI system. As designed, the system is non-deterministic, so we must always carefully assess the responses it provides.\u201d<\/p>\n

Where should a small or medium-sized enterprise begin if it wants to explore AI?<\/strong><\/p>\n

\u201cFirst, it needs to understand it,\u201d Nardi says. \u201cThe distinctions I am making may seem straightforward, but I do not think they are widely understood. Before looking at a specific system, businesses need some basic AI literacy: how these systems work, what they can reasonably expect from them and what they cannot.<\/p>\n

\u201cThen, of course, I believe the most important step for a company is to identify where these systems can add value within its own processes.<\/p>\n

\u201cIf you look at the analyses published so far on AI adoption in businesses, this remains the critical issue. Beyond understanding the principles or how a particular type of system works, the challenge is knowing how to apply it to improve processes or product quality. Many projects aimed at introducing AI systems have been\u2014I would not say failures\u2014but weak precisely in this respect.\u201d<\/p>\n

Looking ahead to the next three years, what concrete objectives within FAIR\u2019s work would help us assess whether we are successfully bringing AI into business and society?<\/strong><\/p>\n

\u201cFAIR\u2019s plan includes a range of actions to support wider adoption of AI systems,\u201d Nardi concludes. \u201cFAIR is particularly focused on research and on developing the tools needed to translate it into products and solutions that can be used in practice.
\n\u201cThis will require us to address the gap that unfortunately exists in Italy more broadly between research and application. In the specific case of AI, I believe one important aspect is the point I raised earlier: helping highly innovative small businesses and startups develop their potential may be one of the most effective approaches.
\n\u201cOne of the instruments FAIR intends to use for this purpose is cascade funding calls for SMEs, underpinned by practical collaboration with universities and research organisations.\u201d<\/p>\n","protected":false},"excerpt":{"rendered":"

An interview with Daniele Nardi, Full Professor of Artificial Intelligence at Sapienza University of Rome and Board Member of the FAIR Foundation Turning Italy\u2019s expertise in artificial intelligence research into applications for businesses and public administration requires training, investment, infrastructure and collaboration between researchers and industry. This transition is the focus of our conversation with…<\/p>\n","protected":false},"author":6,"featured_media":5092,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[46],"tags":[],"class_list":["post-5093","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-news"],"_links":{"self":[{"href":"https:\/\/fondazione-fair.it\/en\/wp-json\/wp\/v2\/posts\/5093","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/fondazione-fair.it\/en\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/fondazione-fair.it\/en\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/fondazione-fair.it\/en\/wp-json\/wp\/v2\/users\/6"}],"replies":[{"embeddable":true,"href":"https:\/\/fondazione-fair.it\/en\/wp-json\/wp\/v2\/comments?post=5093"}],"version-history":[{"count":3,"href":"https:\/\/fondazione-fair.it\/en\/wp-json\/wp\/v2\/posts\/5093\/revisions"}],"predecessor-version":[{"id":5096,"href":"https:\/\/fondazione-fair.it\/en\/wp-json\/wp\/v2\/posts\/5093\/revisions\/5096"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/fondazione-fair.it\/en\/wp-json\/wp\/v2\/media\/5092"}],"wp:attachment":[{"href":"https:\/\/fondazione-fair.it\/en\/wp-json\/wp\/v2\/media?parent=5093"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/fondazione-fair.it\/en\/wp-json\/wp\/v2\/categories?post=5093"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/fondazione-fair.it\/en\/wp-json\/wp\/v2\/tags?post=5093"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}