This article is the second in a series on David Ben-Gurion’s exchanges with Prof. Amos de-Shalit.
A year and a half passed “quietly” since David Ben-Gurion and Prof. Amos de-Shalit last corresponded. During this time no letters or ideas were exchanged between the two. Nevertheless, the subject seems to have continued to preoccupy Ben-Gurion’s thoughts, to the point where he began to read scientific articles by renowned physicists on related subjects.
On June 10, 1959, Ben-Gurion decided to break his silence and sent de-Shalit a short and to-the-point letter. In it, he mentioned that a few days prior he had received a collection of articles written by Niels Bohr, the father of quantum physics, entitled “Atomic Physics and Human Knowledge,” which he had read. Summarising his understanding of the physicist's thinking, as well as his own position, Ben-Gurion quoted Bohr as stating that, “…such a viewpoint condemns as irrelevant any comparison of living organisms with machines.”
Bohr published the letters in 1958 and it included a collection of articles and speeches that he wrote between 1932 and 1957. His articles dealt with philosophical aspects of physics and the many implications of quantum physics on human knowledge and understanding, as well as the interpretation that we can give to various phenomena. Apropos interpretation, one of the reasons that scientific conferences were not held on quantum mechanics during the 1950s was the dispute among physicists on the role of observation and interpretation. Bohr led the school of thought known as the Copenhagen Interpretation, named after the city where he worked in Denmark.
As part of this interpretation, which had been proposed already in 1920, it was claimed that scientific observation had a decisive effect on the outcome of an experiment. To many, this idea was (and still is) counterintuitive if not anthema to human understanding. The basic particles in nature, like the photon, behave and exist at the same time both as a particle and a wave, and the decision which of them is the actualized state of the particle is determined by the observer. In other words, the experiment and the measurements carried out by humans effect the very thing they are observing.
At the end of June Ben-Gurion paid a visit to the Weizmann Institute of Science, where he met with de-Shalit and his most famous invention: WEIZAC, Israel’s national computer. We know about this visit due to another letter Ben-Gurion received, this time from a researcher by the name of Dr. Michael Shimshoni, who wrote: “I listened with great interest to the debate between you and Amos de-Shalit and others near our electronic computer. Because you showed that this subject interests you, I permit myself to offer for your perusal two scientific books dealing with the connection between computers and thought.” One of the books was by John von Neumann, the father of the modern computer, who had died about two years earlier, in 1957. The letter also reveals that Ben-Gurion had claimed during the debate that a computer cannot write a play like Hamlet, and Shimshoni wrote that: “Computers are at a lower rung on the ladder of intelligence.”
Meanwhile, as this fascinating scientific-philosophical discussion was taking place, at the end of June, 1959, a political storm was raging over Israeli weapons sales to West Germany. Ben-Gurion demanded that the ministers who had voted against the deal resign. After they refused, on July 5, Ben-Gurion announced his own resignation (as he would later frequently do).
- When Ben-Gurion pondered AI and the role of ‘thinking machines’
- A city of salt, peace and no Arabs: How a Jewish merchant in the Russian Empire envisioned Israel in 2040
- ‘AI firms destroy jobs – we try to preserve them’
Two days before Ben-Gurion’s resignation, on July 3, 1959, he received a four-page letter from de-Shalit. The letter stated that Niels Bohr had already retracted the statements that Ben-Gurion had mentioned in his letter. De-Shalit clearly elucidated the main problem that we as human beings have when we try to understand and imagine events in the quantum world: According to de-Shalit, because of our inherent inability to observe these, we try to imagine the microscopic world by means of images from the world we know – the macroscopic world. “We are forced to give the electron characteristics that so oppose, ostensibly, the particle and the wave, only because we insist on depicting a system that undoubtedly belongs to the microcosm by means of concepts borrowed from the macrocosm.”
De-Shalit took Ben-Gurion on a fascinating scientific journey in which he explained plainly the scientific developments and the increasingly blurring line between the inanimate and the animate: He described the progress and the breakthroughs in understanding the mechanism of cell division and the ability to create life in a test tube (by means of fertilization or insemination). From there, de-Shalit moved on to the complexity that still exists in understanding the inner workings of human thought, and describes a relevant and groundbreaking experiment undertaken at MIT.
The year of AI
The year 1959 was of the utmost importance for artificial intelligence. That was the year when Prof. Arthur Samuel first coined the term “machine learning,” and defined it as “a field of study that gives computers the ability to learn without being explicitly programmed.” Even now, in 2020, machine learning is the most common means of developing systems of artificial intelligence. The second event in 1959 was the opening of the first AI lab in the United States, at MIT.
De-Shalit described in his letter one of the first AI experiments in the world. MIT researchers tried to teach a machine to visually distinguish between different geometric shapes. De-Shalit described simply the way the experiment was carried out and the way the machine “learned” and “remembered” the information it “saw.” He also explained that the machine had no way of understanding the geometric form, and therefore people had to provide it with feedback as to whether the answers it was providing were correct or not.
“After about 200 ‘lessons’ conducted this way, the machine knew how to differentiate between shapes, even when they were slightly distorted,” de-Shalit wrote. The succinct explanation described the process of information annotation (data labeling), as is done today, when we want to teach a machine to identify some objects, such as a cat or a car. The machine by itself cannot identify a cat – unless it is given pictures of cats in various positions, and it is told that these are cats. De-Shalit’s letter provided an example in which he describes the way he teaches his son about good and bad: “When I think sometimes about the number of times I reiterate to my son what is good and what is not good in our world, I’m not surprised that people (at least some of them…) differentiate between good and bad; I am surprised that they do this so inefficiently.”
After the explanation as to how machine learning works, de-Shalit moves on to explain the difference between a machine and human memory. While in a machine we have to erase and reset the memory every time because of limitations in storage space, with people memory is unlimited in scope and span, and allows us to create and express countless new ideas on countless topics.
De-Shait’s brief explanation encompassed the state of the art of the discussions and concerns about AI at his time and they are still relevant to this day, being pondered by some of the greatest minds in the world. We are at a stage in which AI can carry out discrete tasks very well, going even beyond human ability – what is known as “artificial narrow intelligence.” However, we have not reached the point where AI can learn any subject and carry out any task in different fields, so that it exceeds human abilities – what is known as “artificial general intelligence.”
De-Shalit ends his letter with an apology for its length, and with self-deprecating humor says that “a good machine would certainly have written this whole letter in one or two lines.” Some 60 years later, there is now a machine called GPT-3, which can write and formulate texts on almost any subject and humans can hardly tell that they were generated by a machine.
The day before de-Shalit’s letter arrived, Ben-Gurion was interviewed by Yishayahu Ben-Porath of the Israeli daily newspaper Yedioth Ahronoth. The interview itself dealt with political and diplomatic affairs of the day, but ended on a more personal note regarding Ben-Gurion’s leisure-time activities. Asked about his reading habits, Ben-Gurion answered that he reads at night, not during the day, and that it “depends on when I finish work and it depends on how many hours I sleep. I can’t fall asleep without a sleeping pill, sometimes it works an hour late and sometimes it doesn’t even work that quickly – then I read. The day before yesterday I couldn’t fall asleep for more than a few hours, and so I got up and read.” Ben-Gurion went on to say that he read books about Judaism, ancient Greece, India, and scientific books, noting that he doesn’t read fiction. One of the interesting and relevant points for our purpose is that Ben-Gurion mentioned reading books about philosophical ramifications of modern scientific research.
On July 27, Ben-Gurion sent two letters. The first was to Michael Shimshoni, the WIS researcher who had initially reached out to him, in which he thanked him for recommending the two books a month earlier. Ben-Gurion wrote that he already had von Neumann’s book and had read it, although the first part he couldn’t “read at all, because the mathematical formulas were incomprehensible to me, unfortunately, due to [my] lack of higher mathematical education.”
Ben-Gurion also noted that it was clear to him that “a machine does not think – it carries out physical and mechanical actions only, and there is a vast difference between human thought and the work of a machine.”
The second letter Ben-Gurion wrote was to de-Shalit, responding to the latter’s detailed missive. There he launched into a deeper philosophical discussion about the difference between human thought and the ability to learn, and especially the inherent difference between a human and a computer in terms of awareness to the actual learning process. Ben-Gurion mentioned the experiment at MIT, noting that MIT’s machine and de-Shalit’s are not actually examples of learning machines, but rather of only tools: “in essence [these] are no different than glasses or binoculars or some other equipment that a human being creates with his intellect.’”
From there, Ben-Gurion moves the discussion to the ability of a human and a machine to make decisions autonomously. That is, the ability to control themselves and their actions independently. Ben-Gurion compared a human’s legs to a car and wrote that while “a car can move more quickly and continuously…than a human’s legs, it can’t decide for itself where to go and whether to start or stop its movements, which even a young child, who has his own will can do.”
In 1939, General Motors presented for the first time what a car would look like in 20 years, that is, in 1959. The vision was of a fast, autonomous road system, to which self-driving cars would be directed to and would then travel on. Indeed, in 1958, General Motors already had a number of cars outfitted with sensors which could allow them to receive cues from a future smart road.
Ben-Gurion apparently could not imagine that his letter was in fact describing our reality today, 60 years later, and the action of the autonomous car powered by AI, which can decide for itself where to go. In his letter Ben-Gurion noted that the car should “‘be test run’ at first to make sure it is operating properly, but call this test ‘learning’ – this would be a wrong usage of the word. The car does not know that it is being run, while the human knows they are learning.”
One of the obstacles in autonomous vehicles is the ‘trial run’ to which Ben-Gurion referred: AI systems which independently control cars need to learn to understand the road, traffic laws and driver behavior. That same process of learning proceeds by means of “trial run” of autonomous cars over hundreds of thousands, and even millions, of kilometers of roads and in various situations, and has and is still taking place. Today there are driver assistance systems that can identify and warn drivers of impending dangers due to these test runs.
Ben-Gurion’s opposition to the idea that a machine could “think” the way a humans do was vehement, and in the letter he claimed that a “human can think things that no one thought of before. They put together new combinations and know that they are thinking. You cannot claim this for any machine.”
Nine years earlier, in 1950, Alan Turing opened one of his seminal articles with the question: “Can machines think?” He develops the idea and hones in on another question: "Can machines do what we (as thinking entities) can do?" Turing then offers a test that would soon bear his name but was then called the “imitation game.” In the so-called Turing Test, a person corresponds with a machine. The test is passed by the computer if the person cannot discern that they are speaking to a machine. That test is still a benchmark for AI, though it has undergone changes over the years.
Over the following six months there was no correspondence between Ben-Gurion and de-Shalit, because the country entered a period of elections and a new government was formed (it continued to rule until January 1960). At the end of January de-Shalit sent a letter to Ben-Gurion in which he told him about an article he wrote together with the physicist Alfred Lande. He explained the scientific problems their paper described, noting that the correspondence between the two had encouraged him to write an article about the fundamental ideas of quantum physics. A few days later, Ben-Gurion wrote him a brief reply in which he thanked him for their conversations, which he said he had enjoyed. Ben-Gurion also wrote that he was pleased that their debate had led de-Shalit to write an article, noting the exchange had not been for nothing.
The years in which the two men corresponded were a period of great progress in science and engineering, this was also a period that saw many advances in artificial intelligence and the field’s rise to prominence. From 1957 to 1974, AI entered its “golden age,” riding on the hopes that it could and would change the world as we know it. The combination of huge progress in three fields – AI, quantum physics and nuclear energy – created a new generation of science fiction, for example, that helped stir the imagination of generations of future scientists. In the 1950s, Isaac Asimov published his book “I, Robot” in which he described the laws of robotics (and AI) which are still held to be valid to this day. The technological and engineering advancements of the 1960s led to worldwide debates on the future of labor and “the end of jobs” – the fear that machines can replace us is still present and debated today.
At the height of the global debate about technology and labor, and about 10 years after the correspondence between de-Shalit and Ben-Gurion, the latter wrote a letter in November 1969 to someone else, nothing that there was one solution and it was a “sophisticated technology which would allow one worker that uses advanced technologies to do the work that now requires ten workers or more. In my opinion that will require a complete revolution in all fields of labor, and that depends on the government,” Ben-Gruion wrote.
According to his letters, Ben-Gurion recognized the importance and urgency of the scientific and technological breakthroughs underway at that time. He recognized the huge potential these developments could have for Israel and was more than acutely aware of the threat these technologies could pose if possessed by enemy countries. In his letter to a youth organization, sent on January 16, 1960, he wrote: “We are living now perhaps in the greatest revolutionary period in the world. The developments in science and technology have shortened distances between countries and although imaginary borders still separate people – the sound of the radio goes from one end of the world to another and knows no barriers.”
I want to end with a sentence that Ben-Gurion wrote to de-Shalit in one of his letters about the difference between a human and a machine. It is a difference that still exists, more than 60 years later, and takes on new significance in a world that is powered by artificial intelligence, assuming an ever-growing role in our lives: “Wonder is the beginning of wisdom, rightly wrote Socrates (or Plato – it is difficult to know what the teacher said and what the genius student placed in his mouth) – there is no machine that is capable of wonder.”
Dr. Tomer Simon is the National Technology Officer for Microsoft in Israel