Christos Papadimitriou Speaks on “Artificial Intelligence: its History, its Present, and its Uncertain Future”
Christos Papadimitriou, Donovan Family Professor of Computer Science at Columbia Engineering at Columbia University, USA, and Principal Scientist at the Archimedes Research Unit of the Athena Research Center, Greece, spoke about “Artificial Intelligence: its History, its Present, and its Uncertain Future” during the ten-year anniversary event of diaNEOsis think tank, which took place on March 11, 2026, at the Stavros Niarchos Foundation Cultural Center (SNFCC).
He spoke about the five distinct periods in the history of AI: first, the optimism associated with the theories of Alan Turing, John McCarthy, and Marvin Minsky, and the founding of the discipline of AI during the period 1950-1960; second, the "cold" period of AI in 1970-1980, where the initial optimism of the previous decade was faced with lack of trust, lack of funding, multiple failures, and even ridicule; third, the time period of 1980-1988, where the branch of expert systems showed impressive laboratory results but proved rather fragile in practice; fourth, the historical period of 1990-2000, where AI theories once again faced heavy criticism, ridicule, and very low support; and finally, the fifth period, which began in 2012 and continues today, and can be described as a period of “sudden, unexpected, and unexplainable success.”
However, Christos Papadimitriou argues that four important factors may explain this sudden triumph:
1) Moore's Law - for about seventy years, the processing speed of computers increased by about 60-70% per annum, thus making the application of neural networks technologically feasible in recent years.
2) The Rise of the Internet - the internet generated and provided billions of image and text files for AI models to be trained on.
3) The Relentless Commitment of the Visionary Scientists of the AI Field - scientists such as Geoffrey Hinton, to name just one. A scientist who was recently awarded the Nobel Prize in Physics in 2024 and the ACM A. M. Turing Award in 2018.
4) The Computational Technique of Backpropagation - a simple idea that had been expressed about three decades earlier, was considered “old and tired,” and yet suddenly became one of the core drivers of the explosive growth of AI.
Christos Papadimitriou expresses the view that, according to Aristotle’s definition, AI is not a science. He states that “AI remains an empirical phenomenon that shows [significant] resistance to scientific understanding.” Although there appears to be a “contradiction between empirical triumph and scientific shallowness,” this contradiction defines the character of AI and should provide guidance to researchers on how to approach it. Christos also expresses his “profound respect and admiration” for the achievements of AI and states that “AI is one of the greatest inventions of all time.” He also reminds us that we should not forget that all a Large Language Model (LLM) does is predict the next word based on the words that preceded and the question that it received from us.
In the present time, a fierce race for AI supremacy is taking place between gigantic companies like Google, OpenAI, and Meta, and superpowers like the USA and China. This creates an intensely competitive landscape that leaves little room for a calm, collective, and democratic approach towards AI. Legal regulation is necessary but de facto impossible under the present conditions. Christos also points out that the field of AI has "absorbed the entire scientific heritage of human civilization" and should be thought of as the end product of the entire work of thousands and thousands of dedicated researchers and scientists. There also seem to be no globally coordinated efforts in building this technology; at least not similar to the global initiatives observed when humans were developing nuclear power and were trying to reverse the consequences of climate change. As Cavafy might have said, humanity is developing the highly dangerous technology of AI "with no consideration, no pity, no shame."
Christos Papadimitriou predicts that in the near future all the relatively easy problems to solve with AI will be solved, and that we will experience a slowdown in the growth of AI. He believes, however, that the real problems will remain unsolved. One of the real problems, he says, is the problem of grounding in AI. In simple terms, humans associate words with life experiences and personal moments, whereas, Large Language Models associate words with tens of thousands of other words, through statistical correlation. This represents a significant difference.
In addition to this, nobody can rule out the possibility of achieving artificial general intelligence (AGI), which simply means that machines and robots may be capable of exceeding the performance of humans in all aspects of human life. We must consider the implications of achieving artificial general intelligence urgently.
We must think about the problems that may arise from mass unemployment, rising inequality, and the fact that artificial intelligence seems to be "a mechanism that produces inequality." We must also consider the potential implications of high concentration of power among a handful of states and a handful of companies. Another alarming scenario to consider is that one day the robots may take over.
Finally, there are two future scenarios regarding AI and its environmental impact. The first assumes that AI and its data centers will consume enormous amounts of energy and water and will thus aggravate the already alarming environmental problem. The second scenario assumes that AI may lead to the discovery of new materials, the reduction of carbon pollution, result in optimized energy consumption and be the "magical human creation" that will solve our environmental issues and create a safe living environment for humans.
Christos Papadimitriou concludes that "the only certain thing regarding artificial intelligence is that there will be uncertainty."
Watch the entire speech here (in Greek): https://youtu.be/r52ivtRctX8
