John Henry and artificial intelligence – 05/20/2023 – Candido Bracher

There are composers whose music offers a profound perspective on themselves and their universe. One of my favorites is Johnny Cash.

Cash’s childhood was marked by the arduous daily life of a small cotton plantation in the southern USA, in the difficult period that followed the 1929 crisis. Having overcome countless setbacks in life, many of his songs reveal empathy for men who did not have the same luck.

He sang the story of Ira Hayes, a Native American who appears in the famous portrait of six Marines planting the American flag at the end of the Battle of Iwo Jima and who died an alcoholic ten years later, unable to adjust to life without prospects on the Indian reservation. . Cash also performed at Folsom Prison and San Quentin, dedicating songs to inmates.

One of his most iconic songs, “The Legend of John Henry’s Hammer”, tells the story of a black railroad worker in the late 19th century, who was the fastest and most powerful with his sledgehammer, driving nails into the stone for the tunneling, until the company brought in a steam compressor to carry out the work. John Henry does not conform and challenges the machine to a competition in which he wins, but dies of exhaustion the next day.

I remembered the song a few days ago, when I received in a WhatsApp group a promotional film for a new Microsoft tool, Copilot, which combines the generative artificial intelligence (AI) of GPT-4 with the company’s various applications.

Big parenthesis: Generative AI is a category of artificial intelligence capable of generating new data, such as text, images and tables, similar to the data with which they were trained. Neither the space in this column nor my knowledge allow for a clear explanation of how generative AI works, but I gathered some information that is worth sharing.

1) Generative AI works from “large-scale language models” (LLM). Given a string of words, the model selects the word most likely to be next in the string.

2) Curiosity: in 1913, the Russian mathematician Markov was a precursor of the technique, based on the statistical analysis of the novel “Yevgeny Onyegin”, by Pushkin, considered the father of modern Russian literature.

3) Unlike traditional computer systems, in which the programmer is perfectly aware of the process through which input data generate certain results, LLM systems are not exactly “programmed”, but rather “nurtured”. By adding processing power or data, the behavior of the system is significantly altered, without the programmers knowing how to explain how this happened.

Always emphasizing that the objective is to free the executive from mechanical and repetitive functions, so that he has more time to devote to functions that really require creativity, the film describes how Copilot will be able to perform a series of tasks that I used to consider important parts of the my executive job. Using only “natural language” —that is, the language we speak—, the program will be able, for example, to summarize our virtual meetings, correlating the subjects discussed with our agenda, emails, chats, contacts and documents and to prepare texts and materials aimed at our team, superiors, or customers, containing instructions and information necessary for its fulfillment.

Of course, the final decision always rests with the professional, who may or may not accept the suggestions received, or ask for improvements. But a good part of the initiatives that used to be differentiating factors of executive performance may now be suggested and automated by artificial intelligence. And we all know this is just the beginning. I was not surprised that the song about John Henry, which I had not heard for many years, spontaneously came to mind. The surprise is due to the fact that this time they are not manual workers who see their activities threatened with obsolescence due to technological progress, nor even specific cases, such as silent film actors, or chess champions.

Now the phenomenon is widespread for professionals with “higher education” such as doctors, lawyers, engineers, executives and even writers; A relevant part of the activities performed today by these professionals can be carried out with more quality and speed by the new technology.

Big changes provoke reactions, and it’s been no different with programs like GPT-4. Last month, more than 1,000 AI experts and technology industry leaders signed a manifesto demanding a pause of at least six months in the development of programs based on the new technology, to allow time for the elaboration of a normative framework capable of protecting from “profound risks to society and humanity” posed by generative AI.

Yuval Harari, one of the signatories, emphasizes that “we need to buy time to modernize our 19th century institutions for a world with artificial intelligence; and learn to master it before it dominates us”.

Henry Kissinger joins the chorus: “Artificial intelligence systems with the power of GPT-4 and beyond should not be enmeshed in the lives of billions of people faster than cultures can safely absorb them.”

Not all, however, show the same fear. Steven Pinker, a professor of psychology at Harvard and a renowned student of the workings of the human mind, believes that new technology merely expands the limits of our brains, just as in the 1960s computers expanded our ability to calculate and account, or the ability to seek information in the 1990s. Bill Gates is another who says he does not believe that a pause in development can work globally.

I, who know much less than all of them, have many doubts. Of course, I am scared by the prospect of unrestricted development of systems whose operating mechanism is not fully understood even by their idealizers. On the other hand, I think it is essential that we examine closely whether our resistance to generative AI does not originate mainly in a protectionist attitude that we university students adopt in the face of a technology that threatens the security of our jobs.

We, who accept as tolerable side effects the obsolescence of manual workers like John Henry in the countless cases of automation that we have witnessed in our lifetime, would we be reacting differently when the “victim” is us?