Maturity 2023, the perfect lineup for an Italian topic on artificial intelligence

Ten questions – and as many answers – to understand what artificial intelligence is, especially generative intelligenceand why everyone is talking about it. This lineup of topics can be applied, not necessarily in this order, to a possible track that asks candidates to address the characteristics, advantages, limits and future of AI. For those wishing to further deepen their knowledge of this revolutionary technology, before (or even after) the Italian high school exam 2023can listen “Italian”the Italian Tech podcast that tells it in a simple and exciting way through the interventions of some of our country’s leading experts.

What is Artificial Intelligence?

It is the ability of computers, through sophisticated algorithms, to analyze data, learn from them and then make decisions similar to those that a human being would make. Algorithms, in short, are the real gears of the AI ​​that we use every day. Sometimes without even realizing it. Algorithms are those used by social networks, for example, when they suggest a post that doesn’t come from our contacts. A photo on Instagram, for example.

And there’s an algorithm at the basis of how Spotify works when it offers us a song that we might like and that we may never have heard. This happens because the artificial intelligence has studied our habits and “recommends” us a content similar to what we usually like. It also happens at home, on TV, when Netflix or Prime Video show us films or series that may interest us. But even when Alexa gives us the right answer: it is thanks to the algorithms, in fact, that Amazon’s assistant is able to understand us and “find the words” with which to respond.

Why is there so much talk about artificial intelligence lately?

Because a new form of artificial intelligence, which computer scientists as researchers have actually been working on for years, has suddenly become available to everyone. It happened in November 2022, when a Californian company called OpenAI launched ChatGpt, a chat capable of conversing with its users as naturally as a human being. This ‘new’ AI is defined as “generative” because it is capable of generating texts and images by simulating human reasoning and creativity.

How does Generative AI work?

Generative artificial intelligence, such as ChatGpt, is based on so-called LLMs (Large Language Models). LLMs are deep learning algorithms typically trained on millions of texts from various sources: books, journal articles and websites, for example. In this way the AI ​​is able to understand and in turn generate text in a natural way. There is also a generative artificial intelligence capable of producing images from text. This is what Midjourney does, the best known and most effective model for generating realistic photos or even works of art. The user asking to generate these images does not necessarily have to know how to draw or take a picture. He just has to have a lot of imagination.

What can Generative AI do?

Generative AI can write a story, an unpublished song, a poem. But he can also draw up a training program, or suggest a recipe, or effectively translate a sentence from English into Italian. Furthermore, generative AI is capable of replicating the style of famous writers, photographers or painters. We could ask you, for example, to write a short story as Nick Hornby would or to produce a sonnet as Shakespeare would. The answers provided by tools like ChatGpt are usually consistent and plausible, but don’t trust them blindly. In fact, even generative AI can make mistakes.

What is a Prompt?

It is an English word that is used very often today, when talking about artificial intelligence. It is a “command”, a text instruction. It is the initial sentence or question that is provided by the user to obtain a specific answer or elaboration. “Write a list of video ideas for TikTok,” for example, is a prompt. The richer in detail and context the prompt is, the higher the possibility that the AI ​​- based on this information – will produce a text or an image.

Is ChatGpt Really Smart?

No it is not. ChatGpt doesn’t really understand what it writes. In short, he is not able to think like a human being. ChatGpt looks creative but it’s not. In practice, the underlying algorithm of ChatGpt, after having analyzed millions of texts, does nothing but predict with extreme precision which word – in a given context – has the most possibility of appearing after another. ChatGpt is a powerful, amazing, incredible expression of statistical calculation.

AI, therefore, is not – yet – as creative as a human being. It cannot produce true art, it is not able to trigger the spark that comes from the connection of two sensations, which, meeting each other, turn on a light bulb in the mind, for example.

What are the advantages and limitations of generative artificial intelligence?

Generative AI can make us more productive and can help us automate elementary processes that typically require a lot of time and little application of critical thinking. For this reason, tools such as ChatGpt can prove to be valid assistants in daily activities, both private and at work. The speed with which answers are generated and the immense knowledge on which the machine draws are an undisputed advantage. But there are limits. The texts generated by ChatGpt could contain prejudices – the same ones learned from millions of texts written by man over time – and above all they could be wrong. In fact, generative AI suffers from “data hallucination”. In trying to answer the answers, it can happen that the AI ​​generates plausible but completely false facts. In short, the AI ​​could lie, and well. It is especially risky to ask her for mathematical operations, quotes and scientific arguments. Furthermore, generative AI can be used to easily (and quickly) produce fake content – a popular and sensational example is the photo of Pope Francis with a white duvet – and disinformation capable of jeopardizing, for example, elections and democracy itself.

Is artificial intelligence a danger to humans?

The advent of generative artificial intelligence has aroused fears and concerns about the future of our species. Many fear that AI, if not properly regulated, could spiral out of control and even cause our “extinction”. This is according to a recent open letter released by the Center for Safety AI and signed by more than 350 artificial intelligence experts and developers. In the letter, of only 22 words, the danger of AI is compared to that represented by pandemics and nuclear weapons. However, some believe that these concerns are, for the time being, exaggerated as AI is still far from human-level intelligence. And therefore we would still have time to build or develop an artificial intelligence model capable of thinking like a man and, at the same time, of doing it safely.

Should the school ban ChatGpt?

Initially many schools, especially in America, banned the use of tools like ChatGpt in class. Professors, just as happened when the calculator began to spread in the mid-60s, were – and in some cases still are – worried that students could entrust their homework to ChatGpt and therefore give up studying and application of critical thinking. Yet many schools have retraced their steps, inviting teachers and pupils to use this new technology. Provided you do it responsibly and safely. The use of ChatGpt at school, supervised by teachers, will allow children to learn the advantages and limitations of this technology. And to learn more about a tool that in all likelihood, in the future, will have a huge impact on the work they are going to do and on the job search itself.

Will machines really be as smart as humans someday?

We are unable to answer this question with certainty today. But leading AI experts tell us that yes, one day, machines will probably think and plan actions like humans. Neuroscientists, in particular, argue that this could happen when AI can count on a robotic body that will allow it to interact with the outside world and therefore acquire a phenomenal consciousness.