Let us recall some of them. We communicate via the internet, but we complain about the closure of post offices. We have overcome the two-year period of the pandemic, promising to be willing to do anything to overcome it, and now we are surprised by the arrival of the first of a long series of novelties, disappointed that the calendar has not regressed to 2019, to before the global contagion. Let's come to the topic of the day: ChatGPT, a dialogue platform animated by a digital intelligence trained to respond. Let's get used to interpreting this acronym as a fait accompli and remember its genesis. While the world was in lockdown, computers went to school, reprogrammed to read and memorize everything available on the internet, and then respond to anyone who wants to question them.
Thus ChatGPT was born, comparable to a know-it-all, a classmate ready to talk about everything but who does not know life, unaware of the reliability and consequences of his answers.
This was discussed at the recent conference organized on the Campus Est of the University of Italian Switzerland-USI by Lugano Living Lab-L*3, the department of the municipality of Ceresio dedicated to digital development.
The meeting recorded a predictable "sold out" of the available seats. To direct the works, with the assistance of the computer Andrea Scarinci of L*3, the Professor Luca Maria Gambardella (in the cover photo). Among his current assignments, the latter is Deputy Director for Innovation and Industrial Relations at USI, after 35 years of experience in Lugano at the Dalle Molle Institute for Artificial Intelligence-IDSIA. It is one of the main research centers worldwide and which, it should be remembered, was founded by the will of the late Italian philanthropist Angelo Dalle Molle, inventor of the Cynar liqueur. Let's go back to the future, a ChatGPT.
It is an application intended to learn from your readings, your dialogues, and your mistakes. As a schoolboy, he will go on to a diploma, perhaps a PhD, perhaps to become one of our trusted professionals. Who contributes to its development? Surprise: it's us. ChatGPT is modeled in our image and likeness, but not on our conscience. Unlike an insufferable know-it-all, this application has instead been trained to dialogue and remain docile in responses, as well as ready to modify its style based on the preferences of the interrogator. Want an example? This is how the application looks formal: “Chatbot-GPT has been trained in conversation, providing answers to specific questions and participating in conversations in a consistent and natural way, using a large body of conversations and conversational text. During the training process he analyzed thousands of conversation examples, learned to identify common patterns and generate appropriate responses for a given conversational situation. Bottom line, GPT Chatbot is a probability-based language model that predicts the next word in a conversation, or basic text, based on previous words” and recognizes that the skills and above all the “human beings remain irreplaceable and unrepeatable".
The spread of ChatGPT proceeds at a pace never seen before. According to site surveys I www.statista.coNetflix took 42 months to reach one million users, Airbnb 30, Facebook 10, Spotify 5, Instagram only 75 days, while ChatGPT even 5 days, a simple working week. Let's calm down observed Professor Gambardella: this application is no better than human intelligence. But let's not delude ourselves, because ChatGPT is moving towards deep learning. In short, the know-it-all is no longer limited to reading, remembering and repeating, but begins to behave on the basis of what he has learned, despite being unable to verify the sources or understand the context. For example, during the conference the application was able to comment on the "French Revolution”; she nevertheless replied correctly but unreliably when she was asked about a hypothetical "french rotation”. And we are at the sad notes. First of all, ChatGPT bases its knowledge after having consulted any type of written text, without verifying it. Likewise, he learns from every type of request that is made to him. To complicate matters, it is unknown who, where, and on what legal basis processes the data collected via ChatGPT.
This explains why in Rome the Guarantor for data protection has imposed on the ChatGTP property the blocking of access for all Italian users. And we are at the final warnings. As was the case in our school days, we don't take for granted the assignment that our classmate has handed us, just because he's good at it. When, without any critical review, even the ChatGPT user copies the text of others, he automatically assumes responsibility for any errors contained. This therefore explains why, in terms of privacy management, the user must warn that the ChatGPT text he quotes is offered without guarantee. Furthermore, especially customers of this service based in the European Union-EU, must report the "data processing addendum", i.e. that the news has an "artificial" origin. In the future, it was noted at the Lugano conference, ChatGPT will help develop human creativity. In particular, it will free professionals in every sector from the "block of the blank page", from the panic of having to structure a text.
Humans will only have to enrich the drafts prepared by ChatGPT with contents and emotions. Impossible to predict how long it will take for this application to rival the sensitivity of humans. ChatGPT learns from all requests it receives from users: anytime, anywhere, however. So, to slow down its development, just don't look for ChatGPT on the internet and, even if it's free, don't open its site, read it or use it. But if you have done it or will do it, know that you too will help improve it, educate it, and spread it. Then, tomorrow, don't justify yourself by saying it was all someone else's fault.
