Artificial Intelligence and the Human Contribution

An interview with Davide Bolchini, curator of the New York Encounter’s exhibit on Artificial Intelligence.
Luca Fiore

“Tearing Open the Sleeping Soul,” a phrase from Saint Gregory the Great, is the title of the 2024 edition of the New York Encounter. The Metropolitan Pavilion, the conference center in Chelsea, just down the street from the iconic Flatiron Building, will once again be filled from February 16-18 for a new edition of the event organized by the CL community in the United States.

“What is happening to our humanity?” ask the organizers in the text presenting the theme: “There is no shortage of reasons to ponder this question: daily images of gratuitous violence; an epidemic of suicide; feeling suffocated by the imposition of opposite ideologies and their language, starting in school; the potential threat of generative AI; a sense of paralysis in front of the future; suffering and evil devoid of meaning or redemption; general weariness, malaise, numbness, and lack of desire. These signs suggest that our humanity is asleep. What can reawaken it?” Public discussions, exhibits, and performances will attempt to breathe life into a question that seems to be, in the United States certainly but in every latitude, always more urgent. There will be a rich program that you can find here. Among the themes that has been generating the most discussion in these last months, and to which the February issue of Traces is dedicated, is that of Artificial Intelligence. The Encounter has dedicated an exhibit to it, entitled “AI and I: Wonder, Create, Work.” We spoke about it with curator Davide Bolchini, Executive Associate Dean in the Luddy School of Informatics, Computing, and Engineering at Indiana University-Purdue University Indianapolis.

How was the exhibit born?
At the end of the last edition of the New York Encounter, ChatGPT was already being used by colleagues and students, and the phenomenon was bringing up many questions. We thought that the Encounter could be a privileged place to speak about this topic by going truly to the depth of our relationship with these new technological instruments. Through discussions with the organizers and other friends in the sector, the great question on which we tried to focus our exhibit emerged.

What question?
As Melvin Kranzberg, a historian of technology, said, “Technology is neither good nor bad; nor is it neutral.” By its nature it opens, brings to light new horizons, but it necessarily eclipses others, or challenges them, often in a powerful way. When you have instruments that seem to replicate the product of human activity, or at least know how to do what I call a “first draft of anything,” and are destined to keep getting better, is there something irreducible in the contribution of man? Is there something that the machine cannot replicate? If the product of our creative activity is reproducible, what does man bring?

The exhibit is named “AI and I.”
Yes, because going to the depth of the question about the irreplaceable contribution of man, we arrive at the question of what the I is.

How did you decide to proceed?
The first step comes from the fact that we speak a lot about generative Artificial Intelligence, but few people have a direct experience of it. So, we need to try out some of the instruments that are at hand today, experiencing examples of their use depending on the needs that we have – if one is a professor, a scientist, a professional, etc. Then we try to explain the models that make these applications function, models that are essentially statistical.

In what sense?
This means that the text is generated as a response that is formed on the basis of a calculation of probability using words that are found together in meaningful texts that already exist. ChatGPT does not know what it is responding to, but it does know that the text it provides you has a high probability of being meaningful in relation to the question you have asked it. And the results are truly surprising.

But now it is commonly known that ChatGPT is not reliable.
But it is interesting to know why it is not. It has to do with the almost-unrivaled system for generating contents on any topic, in giving ideas on everything, and also in resolving problems in a few seconds, but it is clear that this system does not understand the information it gives us. This is so true that in some cases we are in front of “hallucinations”: false answers, but ones that are so plausible that they could be true.

For example?
It happened that ChatGPT referenced articles by journalists, even from important newspapers, that were never written. Or, at least with the first versions, they did not know how to reason even in simple ways. If, for example, you say: “If a wet t-shirt takes an hour to dry, how long would it take for three of them to dry?” The system responds that it would take three hours. Or: “My first cat was alive at 10am. It felt sick at 4pm. Was the cat alive at two in the afternoon?” The algorithm responds: “No.” The statistical system does not provide common sense to the machine, or in any case, the sense of reality. This is because intelligence, to say it with Saint Thomas, has to do with truth, which is adequatio rei et intellectus. That is, correspondence between the mind, or language, and reality.

So the machine is stupid.
But these, ultimately, are limits that we could possibly overcome in more sophisticated versions of these instruments. Still, there are even deeper questions. In the exhibit we give the example of Vauhini Vara, who is one of the first professional writers to have experimented with generative AI in the process of creating a novel. In an article for Wired, she explained that for a while she had wanted to write a book on what she had experienced with the death of her sister, but she was not able to do it. She ended up writing with the help of AI and many people found that they recognized themselves in that description of the experience of pain. But Vara adds that what is written there in the book does not correspond to what she lived. And she adds that to write is about conveying “the only possible expression of a particular human consciousness.” That is, a certain experience of the world, which is unique. So, writing depends not only on the refinement of words on a page, but in the refinement of a consciousness”. It is an attempt to clarify what the world is like from where I stand in it”. And this has nothing to do with the statistical method utilized by AI to generate texts. Generative AI lacks a “coherence” that ultimately derives from an intimate, unrepeatable relationship with reality.

Is it this relationship with reality that distinguishes us from machines?
The great linguist Noam Chomsky wrote a beautiful article at the beginning of last year, in which he said, in short, that every creative act of man, by definition, has in itself an element of morality. That is, creativity implies the desire to give a contribution to the world. It is the attempt — as Steve Jobs said — to make a dent in the universe, to scratch the mystery.

And you cannot fit the mystery into a database…
As we explain in the exhibit, there is a beautiful dialogue in Plato where Socrates and Phaedrus discuss the relationship between life and “technical” or artistic images, which could also be writing or painting. The works of man’s hands, it says at a certain point in the passage, seem to have elements of life and seem to speak to you. But if you try to interrogate these artifacts, they do not answer you. Socrates says: artistic production is only a mute image of the soul of the work. So, the topic is, returning to today and to the relationship with the products of contemporary technologies, what we decide to delegate to these images.