The great AI debate: What does the future hold?

The Greek scientific elite of artificial intelligence gathered north of the capital in the suburb of Agia Paraskevi on Monday to discuss how this emerging technology can be harnessed and the transformative changes it brings to everyday life and across scientific fields.

Leading scientists, including from Harvard and MIT, such as Efthimios Kaxiras, Dimitris Bertsimas and Georgia Perakis, as well as scholars from the National Technical University of Athens (NTUA), such as George Stamou and Theodora Varvarigou, are participating in a series of seminars organized by the Hellenic Institute of Advanced Studies (HIAS) at the Demokritos Institute of Informatics and Telecommunications & National Center for Scientific Research. The main objective of the five-day event is to foster collaboration between scientists abroad and in Greece.

“I envision a future where artificial intelligence will become the universal language across all disciplines,” said Bertsimas, an applied mathematician and professor at MIT’s Sloan School of Management, adding that AI has progressed to possess “superhuman intelligence.” He expressed excitement about the potential AI offers, even though the tangible results might take several years to materialize. Bertsimas highlighted the universality of artificial intelligence and its applications in diverse fields ranging from the sciences and engineering to law and medicine.

The MIT professor delved into AI as a prediction tool, offering examples of its practical applications. For instance, in the case of heart disease, factors such as the patient’s gender, age, and the number of cigarettes smoked per day are taken into account. However, people are not mere recipients of AI’s findings, he said; they actively contribute to shaping it.

Regarding medicine, AI can identify which patients receive subpar medical care by analyzing vast amounts of medical data. “How do you evaluate the relationship between quality and cost? The only way is to use data,” explained Bertsimas.

During the first day of the seminar, various examples of AI applications in fields like physics and ship navigation were explored. The seminar underscored the significance of education as the foundational pillar for a society increasingly reliant on artificial intelligence as an auxiliary tool.

‘I don’t believe that artificial intelligence replaces us. It is trained by humans and depends on individuals. The issue is that the people who will use it should not misuse it’

“The world is changing much faster than before,” said Stamou, a professor in the School of Electrical and Computer Engineering at NTUA. The way we think about education needs to evolve, including the introduction of interdisciplinary courses where everyone actively participates, such as AI+Physics or AI+Medicine. Stamou further added: “I cannot envision a future where we rely entirely on robots for everything, nor would I trust a robot instead of a doctor. However, we will have to learn to collaborate. Change is inevitable, so it’s better for Greek society to adapt and prepare for what is to come because the world will progress no matter what happens.”

Garbage in, garbage out

During the first day of the seminar at Demokritos, various applications of AI were discussed, including its use in the field of physics. Professor Efthimios Kaxiras, chair of the Department of Physics, and John Hasbrouck Van Vleck, professor of pure and applied physics at Harvard University, spoke about how AI can contribute to a better understanding of amorphous materials and the design of materials with improved performance, particularly in the context of creating smaller and more efficient batteries for smartphones and cars.

Kaxiras said that AI has proven to be highly beneficial in dealing with complex materials-related challenges. For instance, it has been employed in studying magnetic materials, which are vital in various applications like magnetic resonance imaging, better known as MRI, and maglev trains. His team has been successfully using AI models for 10 years without encountering any significant drawbacks. “The problems we are trying to solve are of a technical nature. The models do not pose ethical dilemmas; they are tools to choose one material over another. In pure science, I cannot imagine any drawbacks; artificial intelligence is an interesting and complex ‘weapon,’ as it were, that provides answers to challenging problems,” he said.

Moving beyond physics, AI’s integration into education was also discussed. At Harvard, Kaxiras has witnessed the systematic adoption of AI tools like ChatGPT in the teaching approach. While some argue for oral exams to prevent over-reliance on AI, Kaxiras sees AI as a means to enhance learning, pushing students to engage in more profound analytical thinking and move away from simplistic questions. He advocates for a shift in educational philosophy, focusing on critical thinking rather than rote memorization. Students should learn to assess the accuracy of AI-generated answers and discern right from wrong.

On the topic of ethical dilemmas related to AI, Perakis, an operations researcher and William F. Pounds Professor of Operations Research and Operations Management at the Sloan School of Management, MIT, associate dean of social and ethical responsibilities of computing (SERC) at the MIT Schwarzman College of Computing, and co-director of the MIT Operations Research Center, will discuss it in Wednesday’s seminar. Perakis acknowledges that AI, like any new technology, has both positive and negative aspects, often evoking fear. However, she encourages students to explore its potential impact, both beneficial and detrimental, on humanity. “Even if you create flawless algorithms and input garbage, they will still produce garbage as the output,” she said.

Over the past academic year, MIT students were given an exercise, and 56 of them wrote papers on the subject within a one-month period. One student, Perakis said, came up with an AI application that gathers data from all online conversations of a person and provides recommendations on how to improve their behavior in human relationships based on that data. “It would be trained,” like all artificial intelligence applications, “by you, but could naturally overcome its limitations,” she explained.

Perakis believes that there is an overemphasis on the negatives of this particular technology, and she advocates for more discussions on its positives. “I am extremely optimistic; you can achieve something truly incredible for so many issues, like medicine and poverty. Our mind is like a computer; if you input wrong data, it will make wrong decisions, and the same applies to AI. Humans can use algorithms for assistance,” Perakis said.

“The technological advancement in this field has already taken its course,” she said, “and we should re-train ourselves to have the tools to use it correctly.” For this reason, her colleagues at SERC provided “lessons” on AI to the Senate and Congress. “Educating the world is crucial,” she said, “to dispel ignorance about artificial intelligence and promote understanding. I don’t believe that artificial intelligence replaces us. It is trained by humans and depends on individuals. The issue is that the people who will use it should not misuse it.”