Since the appearance of ChatGPT and its competitors, panic has spread. Some predict mass unemployment, the failure of education, the dominance of fake news, the collapse of democracy and the end of human history. I think these augurs have seen too many movies. If calls for a moratorium are heeded, it would be the first time in history that a new technological invention has been suspended or canceled.
Fear is always of the unknown and is cured with more information. When one climbs to the top of a mountain and approaches the edge of the cliff, one’s legs tremble. But if he takes a step forward and sees that there is a landing a meter below where he can put his foot, he stops shaking. The more information, the less fear.
It should be noted that the power of artificial intelligence (AI) is enormous but limited. Chatbots access vast amounts of data accumulated on the Internet by Wikipedia, some newspapers, patent websites, and other online publications and select the most likely results. It’s like a Google or Bing type search engine but multiplied. The basic operation is the same as that of mobile phones when we are writing a text and they suggest the next word in the message. But the most likely, that is to say the most common, is usually also the most trivial, superficial and dubious. In chatbot language, clichés, catchphrases and bureaucratic language abound. He also has no sense of humor.
My modest experience includes posing on ChatGPT the questions and exercises from my Political Science textbook. The answers tend to be correct when, for example, it is a question of applying data to a mathematical formula. But when the student is asked to research an election or analyze a document, the chat response is something like, “Sorry, I don’t have access to current events or real-time data,†“Sorry, it’s possible that this term refers to a specific field or context that is not in my field of knowledgeâ€.
Chatbots know what is on the internet, that is, they know a lot of what is known, but they do not know what is not known and cannot guess it. They give probable answers, but they don’t understand, they don’t think, and they can’t discover anything that hasn’t already been discovered. They reproduce descriptions of how things are, but do not give causal explanations of why they are the way they are. They are the negation of creativity. Furthermore, they have been restricted from contributing anything new that might be controversial.
AI is also incapable of reasoning based on moral principles. A chatbot has no emotions or feelings, it is unconscious, unable to form opinions or beliefs. It is therefore useless for making decisions involving value choices or establishing norms of behavior. Faced with intriguing questions, he is ambivalent: “On the one hand, this; on the other hand, the other”, “some say this, but others say something else”…
The biggest unknown that has caused so much panic is whether artificial intelligence can learn on its own and successfully compete with natural human intelligence. But AI learns from statistical patterns in the available data, including errors and falsehoods. Right now, hallucinations abound, as absurd answers and nonsense are called. You can only improve if you receive feedback through evaluations of results, corrections and suggestions. In this case, humans will maintain direction.
Well managed by humans, AI can generate immense benefits. Allow me a little anecdote. Two weeks ago I participated in an international colloquium at the Universidad Nacional Autónoma de México. During my presentation, there were 400 people in the room and, I was told, 2,600 connected online from ten other universities around the country.
Using a black and white square that we call a QR code, 56 questions arrived that, in half a minute, an AI device condensed into four thematic blocks, making them viable to answer.
When this type of practice becomes general, assistants, secretaries, accountants, translators, a lot of mechanical, repetitive and servile work will disappear. They will follow the path of bus and subway cashiers, petrol stations, typists with carbon paper, travel agencies, road maps or queues in front of bank windows to withdraw four duros. Welcome the time and energies earned for complexity, inventiveness, discovery and creativity.
The increase in well-being that artificial intelligence can enable is enormous. It is estimated that labor productivity in some sectors is already multiplying. There are already formidable advances in medicine, both for diagnosis and for new medicines; there will be soon to deal with the climate crisis, the scarcity of natural resources and food. We will learn to discriminate against fake news, just as we learned to distrust advertising and election campaigns. And while some traditional exams may become obsolete, nothing can replace the personal exchange with students in the classroom.
John Keynes predicted that his grandchildren would work only 15 hours a week; it was a correct calculation to live like in his time, but we have continued to work long hours to live much better. When computers came out, some asked: “Do you work less now?”. The answer was: no, I work the same hours, but the result is higher. The same will happen with artificial intelligence.
