ChatGPT. The Content Creation Revolution

ChatGPT is a Generative Pre-Trained Transformer, that is, a text-producing application that can be accessed for free from November 2022 on the openai.com site, together with others such as Dall-E 2, which does not produce texts but images. The emergence of ChatGPT is already revolutionizing the landscape of information transmission and content creation. At the time of writing it has already been joined by other competing GPTs, such as Brand and the new version of Bing, produced by Microsoft. A GPT is not Google and doesn’t work the same way. Does not distinguish information by giving its source; combines them into a paragraph or a short essay written in a correct, though usually rather flat and bureaucratic style. At the moment, ChatGP and the other generative ones are even less up-to-date than Google. The information they use stops at about a year ago. According to what they tell me, and I’m writing in mid-February, Bing still doesn’t know that we are in 2023; indeed, if you remind him, he gets angry and says it’s not true.

It’s important to keep in mind that GPT doesn’t provide information, it’s just a language blender. It has been provided with the texts available on the internet, and not even all of them, because it is not (yet) possible, and it uses the syntactic structures of the language on which it is based. The resulting text is not based on the meaning it has, but only on the probability that a certain phrase will be followed by another, with all the errors that can derive from it. Whatever you ask it, the answer will be a string of words that have a good or bad chance of being relevant to the question. If you ask him what sources he used, the result will be a plausible but perhaps non-existent bibliography.

But since things change quickly, the elicit.org site is already available, a search engine capable of instantly finding a bibliography of scientific articles available on the internet for every (or almost) question that is asked. Academia.edu has already existed for years, a very useful site, but it is limited to texts that have been voluntarily uploaded there by their authors, while elicit.org searches every publicly accessible database, saving the researcher the trouble of reading the entire article and immediately provides him with the quote he needs, or so we hope.

There is no doubt that the prose of ChatGPT, as yet not exactly exciting, will soon be perfected. When digital computers are replaced on a large scale by quantum computers, for now existing in prototypes of limited numbers, even artificially generated writing could undergo a quantum leap that will make it immensely more capable. It could also develop, if not a full awareness, also an autonomous beauty of its own, still lacking for the moment.

It will (perhaps) enhance human thought, just as an aeroplane enhances the idea of ​​flight that we derive from observing birds, but despite the panic that the appearance of ChatGPT has aroused, especially among elementary and middle school teachers, it is difficult to think that will replace it. Not only because at the moment it is “only” an algorithm that devours, metabolizes, recombines and spits out billions of sentences conceived by human beings, but for two essential reasons. The first is that human thought is not just writing (it cannot be reduced to a series of statements placed linearly on support); the second is that writing has always been a machine.

Alphabetical writing, in its evolution from cuneiform characters to the Phoenician-Greek alphabet, has always to a certain extent “thought” for those who used it, it has always made us develop some concepts, leading us in one direction instead of another. Without a widespread use of the phonetic alphabet and its discrete components (and therefore already “binary”, or one letter or another) it would not have been possible to conceptualize notions/oppositions such as “being”, “nothing”, “principle of not contradiction” or “principle of the excluded middle”, which in fact are not given in Asian philosophical traditions or are expressed in a completely different way. Try translating the Japanese term mu into a Western language and you won’t be able to decide whether it means empty or nothing, both or neither.

Is artificial writing, which is the next revolution we will have to deal with, indeed we are already doing it, therefore a good, a bad thing, or a neutrality waiting to be used in one sense or another? It is bad, to stay with Plato’s Phaedrus, if seen through the eyes of Pharaoh Thamus, who judged Theuth’s invention of writing as an unfortunate tool that would lead humans to forgetfulness. It is bad if seen through the eyes of Socrates, who preferred conversation to writing. It’s a bad thing if we think that the invention of movable type printing in 1455, which made it possible for anyone who could read to interpret the Bible in his or her own way, helped to foment more than a hundred years of religious warfare. It’s a bad thing if we think that the rise of social media, which began around the mid-1990s but exploded only about fifteen years later, has generated mass narcissism and a cult of individual identity (as well as the desperation of not having enough identity) which is unprecedented in history and whose dangerous effects we experience every day.

It is a good thing if we look back, we weigh the damages that writing understood as a machine (and therefore always already “artificial”) has produced, we compare them with the democratization of ideas that that same artificialization has made possible, and we note that never and then we would never want to go back to the time when writing was the exclusive prerogative of the Egyptian priestly castes or of the mandarins of the Celestial Empire. But neutral it certainly isn’t and won’t be, because it isn’t there waiting to be used. It’s already being used relentlessly, whether we know it or not, whenever we allow our phone to complete a sentence for us. But we also don’t want writing to be handled solely by the new mandarins, the ones who stay on the top floors of Silicon Valley offices. And here the problem becomes political.

The immutable law of technology, however, is that its gifts cannot be refused. Try it, and they’ll chase you to the ends of the earth. While we try to understand in which direction this new gift is taking us, the only task we can assign to Minerva’s noctule is to push its flight deep into the past and trace archaeology of artificial intelligence, in order to understand human beings have always created and worked with bits of intelligence external to them, perhaps ever since ancient astronomers connected stars and planets with imaginary lines, called them constellations and attributed to them the power to influence events on Earth. With minds without brains, literally, but not of negligible power.

And if the appearance of ChatGPT will lead to the end of the school “temino”, personally I will not mind too much. University students, know very well how to write a fake essay, they certainly don’t need Artificial Intelligence to teach them. It is enough to use the amplificatio method, invented in the Jesuit schools in the 1600s. It is a question of repeating the question posed by the teacher in other words, in the form of an answer, piling up sentences upon sentences which have the sole intention of reiterating that it is that’s right, that things are as you said, Mr Professor and Mrs Professor. How many hundreds of “terms” have I read that did nothing but repeat what I had asked for in a more flowery way? Never that anyone said to me: no, I don’t agree at all. For that alone, he deserved the best mark.

And then ChatGPT knows how to use punctuation, which for some of my colleagues is an unknown entity. It wouldn’t be a bad idea to assign them a question, show them how ChatGPT answers and then ask them to improve the answer, adding what they think and above all what they feel, given that ChatGPT knows, but does not perceive the emotional impact of what it knows, while human knowledge is never abstract, it is always a feeling of knowledge. (In any case, the GPTZero site is already available and seems to be able to check whether a text submitted to it has been produced by a generative application).