Artificial Intelligence is advancing at an extraordinary pace. This is also true in Italy, where more and more companies recognize that Generative AI is not just a technological tool, but a distinctive element with the potential to redefine the competitive landscape.

And the promises are significant, as it represents an opportunity for companies, public administration, and workers, as recently highlighted during a press conference by Cristina Conti, Customer Engineering Manager of Google Cloud Italia: “According to a recent Google study, 58% of people could benefit from the adoption of Artificial Intelligence to simplify work, accelerate the performance of certain activities, increase productivity, and have support for creative tasks. But the country system also benefits, because according to our surveys, in the next ten years, the accelerated adoption of GenAI could lead to GDP growth between 150 and 170 billion, which is equivalent to 8%. A delay of even just 5 years would reduce this percentage to 2%.”

And Google is playing a leading role in this new wave of innovation and “Agentic Era”, bringing solutions to the market that take into account the needs of customers and partners. “It’s no coincidence that new products, features, and characteristics are constantly being launched,” reiterated Paolo Spreafico, Director of Customer Engineering for Google Cloud Italy. And the latest announcement from Google concerns Gemini 2.0, the new Artificial Intelligence model, with advancements in multimodality, such as native image and audio output and the native use of tools that lay the foundation for agency experiences capable of planning, remembering, and acting on behalf of the user with their guidance.

As Spreafico recalled, “this is an experimental version available in the form of APIs within the AI Studio platform. Essentially, the real novelty of Gemini 2.0 concerns the perception of the surrounding environment and context in multimodality, because it aims to replicate multi-sensoriality, typical of human beings, who hear, see, and smell. The multimodality of Gemini 2.0 summarizes all these characteristics in real-time and has already led to the realization of projects such as Astra – the AI agent built through this model that perceives the surrounding reality and is able to suggest to the end user what is useful at that moment, from a productive perspective – and Mariner – a research prototype that explores the future of human-agent interaction, starting from the browser.”

Google’s strategy is articulated on multiple levels, complementary and indispensable, which allow embracing solutions based on Artificial Intelligence. First, there’s the infrastructure, with the announcement of the new Tensor Processing Unit (TPU) Trillium which “adds computational power three times greater than the previous generation of TPUs, with the same investment. And for us, 2024 was also the year of the diffusion of the AI Hyper Computer architecture in our now 40 Google Cloud Regions, which is at the basis of infrastructure optimization to make Artificial Intelligence work in an optimized and more effective way.”

Another important piece that the company is working on are the models, “and this doesn’t just mean Gemini 2.0, which in its experimental Flash version will enrich the AI Studios. Our goal is to have a platform that is as open as possible, capable of hosting third-party models as well. In the case of Agents, for example, the Google platform, with its ‘Studio Agent’, provides some ‘pre-built’ ones but also gives the possibility to create new ones in a simple way to respond to various business or personal needs. And this is the case of the Italian Agent recently released in Google Workspace, which allows implementing a process that significantly increases productivity: 75% of our customers in this area recognize that the introduction of this feature allows creating content in a collaborative, more efficient, and more performing way.”

Among the models, those for image creation released in 2024 also stand out, including Imagen 3, which starts from a prompt or in a multimodal way from another image, and Vio, the set of APIs for video generation that generates high-resolution outputs starting from other multimedia content.