I think this class of algorithms can be much more important than the transformation of the internet and web. At least what we see, especially when we see the progression of new technologies, that is the quantity of technology startups that we have today and the quantity of technology that we have produced is much more than the first phase of internet. This is a sign that we can say that we are creating an economy potentially explosive and that transform any other thing. Welcome, I'm Emiliano Udizio from UART Italia and this is a new episode of The Big Interview, an episode made in collaboration with Reply. Today we will talk about artificial intelligence between intelligent and intelligent, improved and advanced robotics. We will try to explore how these technologies are changing and changing the way we create, we work, and we will explore our lives. For exploring these fields, we have with us a guest, Filippo Rizzante, Chief Technology Officer of Reply, which will help us to understand the challenges, the opportunities and the responsibilities of this new digital ecosystem. Welcome Filippo and thank you for being with us. Thank you Emiliano, thank you and good morning. You are recently the protagonist of Reply Exchange, which is a little event in which you can tell what is your vision of innovation and the future. You have talked a lot about the event that you have and the revolution in atto. So I would like to start from here. What are the main elements of innovation on which your vision of the future is focused? I don't love the term artificial intelligence, because... not because it is wrong, but because in the course of the years, as human beings, we have attached a lot of fantasy with literature, and with films, TV and so on. In reality, the big progress that there was a progression of algorithmic, a progression of software, a progression of science that I studied because I'm an engineering informatic. So today when we talk about artificial intelligence, which is the main data of technology, of transformation technological, according to the progression of informatics, we talk about class of algorithm very specific, which are called algorithm generative, which we use every day with CHGPT or Perplexity or other applications of this type, which allow, to transform that input in an output of the same type, so from text to text, or in output of different types, for example text to image, So, doing great things that we have experienced through applications like CiaGPT. These technologies today have a lot of potential within the companies, both in the transformation of the processes, as in the transformation of how companies offer their products, or even in their products physically, and then transform much of what is today the business in general, of any industry industry. I can ask you, although you don't like the definition of artificial intelligence, which is also a big debate, which is also a definition that is made a little by chance, but without entering this anecdote, what are the advantages that this type of technology, this type of technological advance can bring, especially in sectors that are very verticalized, like automotive, retail, and the same media as we can. Can you give a panoramica of those that are innovative and profound innovations? Yes, today we have at least five big areas of impact for any company. In particular, the first big impact is on the individual productivity. We already, using GPT or the co-pilot mechanism, for example Microsoft, or intelligent systems that are put on any software, we increase our capacity to produce data, information, or to develop data and information. If you think about it, this increase of productivity is already coming in the last 50 years. Because just think about how an office was, or how an office was, for example, a group editorial like yours 50 or 60 years ago, without digital and without everything we use today through computer and informatics. The artificial intelligence, the part of the generative and applications like CiaGPT or the mechanism of Copal will further increase this productivity. So we will not even know what we will be able to produce and elaborate data and information. This is the first impact. The second impact is creativity. The creativity today has new tools available for example creating images, video, and these new tools They already see, through some examples, they also give new forms of creativity. So, the intuitive human being that, thanks to these new instruments, can produce new forms of expression. The live coding, for example, is the sector. The guys who do performance writing a live code, even maybe produced in part by the machine, combined with a part visual and a part of music that is generated in real time in function of how the programmer is programming. And they become the right performance. These are examples of how creativity and how the algorithm is generated in effect are new machines, new photographs, or new instruments that will be able to create new forms. Then there is the other element, there is the physical part, there is the robotics, the fact that you have to go to automotive, just to go to San Francisco, I have the chance to go there two or three times all year, and this year, in April, the last time I was gone, the quantity of robot Weymo, robot taxi Weymo, that were present in the city because a few months ago it was not so. Today there are everywhere the taxi robot. So the part, for example, self-driving, completely autonomous. We do projects on other machines, like for example drones, which are the other machines for sure more ready for being completely autonomous in the physical world, which is very complex. They use the same technologies, the same innovative technologies, today, especially for the visual comprehension. So this makes progress even in the physical world. Another thing that many people realize, maybe a bit more of you who are editor, is that we have spent the last 20-30 years investing in an digital economy, on the web, which is at least at least at some very clear pg view adwords and we are already destroying that economy because the transition towards applications like cagpt or perplexity that give us already information behind it is different from research through google where the first thing is click and then generate page view in these days we talk about the launch of new browser and the impact that it can have on those instruments that could enter into crisis that are used to now. And, just to say, I read in these days, a interesting statistic on the increase of hours of use of robotaxi. So it's been, not only the number of cities covered, we have some areas in the United States, but also the number of hours of use by the user, so it's starting to be a sort of relationship based on the trust, because that is an element very important. I want to ask you a few things. I've seen a few statistics, obviously when we talk about artificial intelligence and generative models, the economic prospects are all extremely positive. There are various numbers but there are a huge increase in the future, an use is always more massic. I would ask if there is a key to this acceleration and if we can talk about a change of paradigm that perhaps can be more important than what happened when the cloud is. I think that this class of algorithms has the potential to be much more impactful than what was the transformation of the internet and web At least what we see especially when we see the progression of new technologies that is the quantity of startups that are now and the quantity of technologies that are produced is much more than the first phase of the internet. This is a sign that we can say that we are creating an economy potentially explosive and that transforms anything. From the point of view of the impact on the work, there will surely be impact on the work. I'm not worried about the amount total of work, I'm worried about the speed of transformation, because things happen more quickly, or at least so we believe. In reality, many times, we can see the proclamations that were made self-driving, a few years ago, especially when it is about the physical world, they are more lento than what we expect. But in the digital world it is not so. So some transformation can be too fast because some classes of laborers will be compared to other skills or other tools or other things that need to learn. Even social systems, I would say. I would like to talk to another aspect, then I would like to talk to the creativity. We should also change the mental approach. When we think about AI and AI, or complex architecture where more agents work together, You have developed different projects, I remember DevBot, SmartBean, which are some projects that you can tell more in detail. The thing I find interesting is that these complex systems need a capacity of prevision and management of potential errors very refined. In the sense that, because they are complex and complex systems, a small error or a small bug can become very complex to manage within an ecosystem so articulating. tell us what your approach is, starting from some of the projects that you have worked. The mechanism of agents is an evolution of what was done a few years ago with what was called robotic process automation, which was created by bots that were very deterministic, very rules, and on many processes, to define all the possible rules and all the possible conditions and you never finish. So the projects of automation process in large companies were able to automate only one very deterministic part of the process. Now, the agents, in reality, starting from language models, are for their nature probabilistic, They are not deterministic. They have, as you say, a number of errors. This error can be limited if you can make it so that the algorithm, the neural network, the data that has inside the model that people use, is more contextualized possible. So, typically the activities that we do are to go inside the model, with a technique called fine tuning, so to tune the model, as compared to the competencies that we have, compared to the specific process on which people have to act. This makes it so that in terms of results and in terms of error, error, the error is significantly lower and the result is more high. By doing this type of activity, the agent becomes like specialists, specialists of the single activity that they have to automatize or specialists of that knowledge very specific knowledge. It is like to say, all the organizations, the same condeness, within their own language, in their own way of doing things, If we can't help this knowledge and competence within these models, then the agents that create these models can be more experts in the single reality that they are applied, which is exactly the activity that we do. I find this interesting because it brings necessarily this element of variability and of continuous adaptation to the context, which is a central word when we talk about these models, they call it, in the context of the human. And so there is that relationship human-machine that becomes, in some way, that is transform. We have to change our way to approach ourselves. We have to change our way to approach ourselves. We have to change our way to approach ourselves. I don't think so much in the automation complete, but I think more in the fact of making people more effective from the point of view of the execution of the work and then the tempest that are required for a single task, but also in the quality of output. This will allow the organizations to be able to do much more. So not necessarily one should think that the organization or the company should do the same things simply with less resources. In reality, with the same resources can do more and more things. And the same happens for the software. Because the first application on the agent, to build systems, and on the same, on the engineers, we have systems of agents that produce software, that help our developers, our testers, our analysts to produce more efficiently software. But this doesn't mean necessarily that you don't have the people that coordinate the capacity and the thinking to project and to create an application. Maybe it will be less to do a single application, but this will liberate resources for more applications, more software. Today, by the way, that is a pretty clear limit. I've tried to do a calculation very simple with GPT, which you can replicate, which is how many lines of software can be written in the world and how many people are now able to write software. This progression is a line that is going on from the 60's where there were about a million people who wrote thousands of lines of code all year, until a few years ago we can find out between 30 and 40 million people who are now able to write software, informatici o attinenti all informatica, e producono in media 54.000 righe di codice. Questo ti dà il totale del software che è possibile creare. Questa linea sta diventando esponenziale, quindi domani in realtà la quantità di software che riusciremo a creare è esponenzialmente maggiore, ma questo ci darà la capacità ad esempio di realizzare quello che dicevo prima sulla robotica, perché pensare di avere ad esempio robot che effettivamente are able to act in a physical world like this, which is very complex, means that they have a lot of software inside. And from this point, you can go deeper into what becomes a central element of creativity, which we mentioned earlier, which is a central element of your company, the way you operate your agenda, the creativity is a little for a asset that you work, and I would like to define it so. And I would like to talk and talk a little bit about how you have this topic, maybe starting from two projects, one more related to music and the other video, which we have worked recently in these two contexts, how you relate this activity with the discourse that we have done now? In general, we have in Replaybian different ways to stimulate the creativity and the imprenditoriality within our group. So, in the 16.000 members of the group we have some operative structures that bottom-up stimolino these two elements. It is a structure called Social Network and uses different mechanisms, such as knowledge sharing, such as creating events, such as creating laboratories where our people have tools, have skills to be available in order to create creativity and fusion elements. This team is also responsible of what we call challenges. We create ecosystems around our company. So, in several years, we have a challenge on coding, on cybersecurity, on investment and on creativity. The Coding Challenge, which is probably the most participatory, has about 20-30.000 participants in the year, who come from all the world, They typically win, they have 24 hours in which they have a goal and they have to maximize with an algorithm and the platform elabora directly the solution. They participate in teams from all the world, from MIT, from Cina, to the Chinese universities, as a team of professionals. We do it for creating a system that is all around us. So it's all for a reply, an investment of reply, where typically comes an award, all the mechanisms of social content, the various editions of the team on schools and universities, and on the profession We also had a creative challenge typically put brands that were expressing their challenges important brands which the creators of the world then tried to give their answers which in the last few years we have transformed into the AI Film Festival and this year also into the AI Music Contest. The AI Film Festival started four years ago, the first two editions were intern and group reply, So we were only the guys who proposed and proposed and proposed were generated by AI. And the last two editions, thanks to sponsorship and partnership with Mastercard, were we were instead of the Festival of Venice. While the music contest was made with the participation of the Kappa and was live at Kappa Future Festival of Torino. The festival of electronic music. Yes, the sixth world. Ormai ho visto la classifica, molto orgoglioso della città perché K effettivamente ha superato il Coacella, che è il settimo. So, the thing is interesting, that even in these events, that contest of coding has made me think that Netflix has done school in this point of view. All the development of the algorithm of Netflix, they were just from these challenges. Now, when we bring them into a festival, it's open to everyone. All they could then go through this contest, even then on the stage, because then they ended up in a performance. I have 5 finalists of the music contest, which obviously are the director of the K, which I was a little bit of a quality, in reality the quality of the artists is very high, they were surprised by the output, and I would say that, and typically, obviously, they are all semi-professionists, even those who have participated. They use the AI not only for the part of the musical generation, but because in the electronic music it also combines a lot of visual parts. So in reality, it's made of visual, of luci, which then, on the stage, obviously, they make it very well, as we know from the main exponent italian, that is Matteo Miller, an anima, bravissimo, tra l'altro, and that is also represented in KK. So this combination is very linked to the algorithmic, very linked to these generative forms, and so it combines perfectly. So there are artists who were great, who were on the stage, and we will see in the next edition. In the part of the Film Festival, I am in jury, and the president of the jury is Muccino, I saw that there is a part of judges quite interesting. Yes, yes, also on the music contest DJ Albertini, then there was a series of DJs that I know of the least, but international. What I see on the film festival, because I'm in jury, I'm examining the 50 finalists of more than 2000 artists that we received from all the world. I think there are 50 countries that I think there are only Italian Italian, one or two, and I see a lot of progression. One of the reasons why we have made these external events is not only for the network of professionals, which we can use, but also because we can see that we have a very good observation on the progression of technology. And on the video and audio, we see that we are getting to the capacity to be on the screen. You have called this project that you told me, a record of Wired, and also this desire and pleasure in observing the evolution of the great cultural and social and economic related to the technological evolution. In 2014 we organized the first festival of ultra-corting, which was a festival of films in 15 seconds. So that trend is to move a lot of our communication on the video world in a more accurate way, which is a trend that continues to evolve. And maybe it could be even a spunt for the next edition. And the other thing that you made me think about what you said is a spunt for the next question. When we talk about creativity and AI, obviously there are issues related to the authority that are central. A giugno, a judge, in a case called Anthropic, he gave a certain debate because he gave a reason to Anthropic regarding the use of books, in this case, of texts, which were part of copyright for the training of the model, while creating a database, a repository of content, which was protected by copyright, but in some way he gave it to the authorization, at least for training. So there is a big discussion in the world, there is still a synthesis. When you develop solutions of this type, how support your clients and companies in which you collaborate in these aspects? I don't think that the problem is the instrument, but the problem is always how it uses the instrument and what is the output. in particular the generation of images, video and audio, in the end if I generate an image of Spider-Man, and I cite Spider-Man because Disney has used it as an example, the problem is in the person who has used it to try to make a clone not authorized by an intellectual property that is Disney. So the problem is in the output, even in the data that are inside the instrument. It's exactly like saying, I'm listening to music on Spotify, I'm an artist, I'm copying and I'm doing a clone. So, the data are inside me, not inside the model, but the same thing. The problem is the output. So we have obviously a group that is specialised compared to the normative that change from country to country It's all very fluid, as we know, then Europe, AI Act, etc. And we have a group of ethical, as I reply, from about two years ago, a group on ethical AI, So, of adoption internally, but also of conduct towards clients of how these instruments can and should be used. It's obvious that this means that there are also a series of impacts that are well out of the context of creativity. So, this means that, since we move on the physical world, The specific problem of AI when you go on robotics is much more important than, for example, to talk about intellectual property of a generated content. So there are different degrees from my point of view of how these technologies can be impacted both for organizations and for society. that organizations like ours have to take into account when they approach the client. Another thing that seems interesting to me is that there is not only an aspect normative, etic, related to the respect in this case of those that may be the right authority, but then when we talk about virtual assistants, and you are occupying, I think for example a project called Futura, there are other parts that are assistants that are related to people, So there is an approach, as I say, empatical, that is somehow to take into account of an emotional and relational aspect that is not banal. So, from the side, I would like to tell you how you feel and how you feel about the emotions with artificial intelligence. All right, that is an area that we call digital human. We are doing digital human from the point of view visual, then the components are different. You have the brain and the output, the audio, the voice and so on. And obviously the sensor and digital human can also be atterrified on the physical world. So the physical world means that maybe on a screen they have a front camera and microphones, and they perceive the context through that device. We are working with digital human beings that are clearly fake. If you use the MetaHuman of Unreal, which is a mechanism of creating digital human 3D, you can see that it is a puppet, so as you can use mechanisms generative deepfake so that the digital human is over a screen that is a real person. Many times the choice, as in the future, is to not confuse people and therefore to use a visual part that clearly makes it clear that it is a machine. So when you use a person of fantasy or a person of fantasy, cartoon all'interno del digital human. Quello che stiamo poi aggiungendo è tutta la parte sia empatica sia la parte, altre due componenti che sono la parte di proattività e la parte di gestire non l con una singola persona ma con pi persone perch la parte empatica fatta sia from the point of view visual from the point of view of the argument from the point of view from the voice from the view of the empathetic understanding of the person who is interacting with digital human We created a model that is actually a model of agents based on the inside out of Disney, where each agent represents an emotion. There is a paper that has been published, because we have done it using the students who come from the University of Genova who have used it as mechanisms to make their thesis and their scientific part. This model does not mean that the digital human has done it from the point of view emotive. So, given a personal personality, just described, from the digital human, that gives, in quotation marks, then the various emotions, and the input, and based on personal personality, decides which emotion, mainly, and based on the dialogue, and in that case, the digital human is the only, which decides how to respond to the person. And this, through components very sophisticated, ad esempio di invidia e così via poi lo realizziamo anche dal punto di vista visivo quindi se devi esprimere visualmente la rabbia è tutto automatico cioè non è pre programmato quindi se devi esprimere vocalmente cambiare il tono di voce anche lì ci sono componenti ed è tutto automatizzato per cui digital non da solo decide l'emozione in base alla sua personalità a base al contesto del dialogo e and only express it, visually and visual. We are arrived in the chiusura of this journey, we have to have many aspects, I would still want to focus on an aspect that is related to creativity, but also to cultural aspects, because just a few days ago there was a report annual of FederCulture, and the president is cancelled, he gave some data, and, in addition to what said in the past, some ministries with culture and culture, that with every euro invested in cultural operations, there are nine in the world. We've already talked about some of your collaborations with cultural projects, cultural manifestations, music, cinema, but there are also other aspects in which technology is changing the relationship with culture and creativity. And I think about the work you've done for the Jubilee on San Pietro, which also opens scenarios from the world's point of view. It's very different work. Yes, that is a work that we've done, thanks to Microsoft, started in fact from the United States, so with our part of the American company, reply in the United States, which has, for the Vatican, we and Microsoft, digitalized San Pietro, to all the effect, building the digital twin of the entire Basilica. The difficulty of the project is the total respect and precision that we have to have when we create a digital twin 3D in the Basilica of San Pietro. So every element must be perfect and every element must be inserted in a context that is what that monument represents. So we have created those elements, not only the experience, which will be digital twin digital, which represents the historical archive of the Basilica of San Pietro. By the Jubilee, it is not the only one we have done. We have done also another interesting project, which is going live in Piazza Castello, a Torino, which was the digitalization of the Sindone. So it was exposed to the Sindone on a gigantic table, together with the Curia of Torino, because the Sindone could not be exposed, and so it was done this exhibition for the Jubilee completely digital, which had many visitors, and of whom we were being a partner technology partner. This also opens new possibilities of fruiting and of relation to what are the cultural patrimony, which, returning to the work, generates new possibilities, not only in the world of work, but also in how we can do tasks that before were not imaginable for us. Yes, to do more with the same resources, to do more things, which cost even less. Think about it again. Many think about production, for example, return to film, that this will destroy the big house of production. Yes, but it will also be a context to many people who will produce to a quality that they will not reach today, and with their monetization, being a team much smaller, will be able to kill a lot of small niches. This will be an economy much more frammentable, much more democratized, and the current and current, there are always changes in this type. So, I think, the opportunities that the world of artificial intelligence and the generative are much more than the thoughts of the current world that is changed, and therefore the negative impact of transformation. You have used the term democratization. I want to end with a question more related to the industry industry and product. There are also here data that tell how the AI, in their various forms, dimensions and applications, is entering more in the processes and instruments of companies. But in other words, this narrative sometimes comes back to some context, and maybe Italy is a interesting place, which is the accessibility of technologies, which is only the scale of companies that can use these technologies. Is it a narrative concreta, or is it accessible to all these technologies? I'm wrong, honestly, this is already brought to the internet, the democratization of the knowledge and information, except for a small postilla that I will do at the end, but today many of these technologies are always available. It's not that GPT, DeepSeek, Anthropic, or Lama are not available, they are absolutely available. Many times open source, many times, even a low cost or accessible when they are not open source. It is a little less on other technologies, for example robotics. Obviously when you enter the physical world it is a little less accessible. So many projects, for example, humanoid robotics today are very closed, or are available in areas of geographic areas. Another thing is the AI Act, because all the implementation that has put Europe makes sure that there is always a retard of disponibility, especially of new models or of some technologies, compared to the United States or compared to other geographical areas that have more accessible regulations and therefore less limitable. So we see that the vendor, in terms of the interest of making their technologies available for Europe, today are afraid of implications. I see some borders that European Union, in terms of multitudes, try to give to the technological players when they don't respect, from the point of view of the reglementation, some of the European Union. We have talked about this week, a letter from many European companies maybe a discussion of some parts of the law in an eye to a greater access and greater possibility of competing with other actors. So it is a very open question that we monitor. I think for you also also in the central area. Yes, for me, from another point of view, I hope that some mechanisms of chiosura, which are typically linked to the current panorama geopolitical, make sure that Europe invest more on itself, on technology, which is a technology that is purely European. And on this, I think it's an augur that we have all. Filippo, thank you for having brought us in the heart of these new frontier technologies. It's really a question. So, what is the mental approach that will be the difference in this technological change? Individual, to be curious, to be passionate, to study, to learn, to experiment, because the technology is going very fast, it gives many opportunities, so we don't need to be fossilized on ideas, but to be open and curious. and it's what anima the big interview that this episode arrives at the conclusion and then it's best in the way thank you for being with us for who see and who listen the big interview continues on all the big interviews to see you next time bye thank you
