Whoever regularly reads this supplement, has almost certainly noticed the special interest we show in the wonders of AI, and especially with the image and text motors that have made this whole business into something extremely accessible – or perhaps alarming, depending on your point of view. It was actually quite exciting to discover that the office of the person in charge of AI In Israel, in the flesh, is located in the office block right opposite the Mekor Rishon head office. In contrast to the mysterious "person in charge of the internet" whom the IDF Army Radio broadcaster, Razi Barkai, searched for back in 1996, the fact that the man in charge of AI is a real living soul was really moving in itself.
Dr. Ziv Katzir, the head of the TELEM (The National Infrastructure Forum for Research and Development) program for Artificial Intelligence at the Israel Innovation Authority, is in charge of the national AI infrastructure – "its integration" and "management", as per the definition of his tasks in the usual government-corporate-management jargon. Katzir dedicated his doctorate to the question of what happens to AI systems when people try to cheat them; so it might be a good thing that we met him momentarily before we came to understand that AI will be replacing us as writers and editors.
Q: Maybe we do need to learn how to cheat it?
"Why? "Generative AI will not make us redundant," he concludes at the offices of the Israel Innovation Authority in Jerusalem. "Some professions will disappear and some will be born, that's apparently true, but there will be no mass unemployment."
Q: There won't be?
"The problem always applies to the 'desert generation'. There will always be people who find it difficult to adjust to new technology, and we simply need to help them. Over the course of time, we have never really witnessed a technological revolution that has resulted in large waves of unemployment."
Q: Prime Minister Benjamin Netanyahu thinks otherwise? Prior to the current war, he spoke often about AI, voicing concern over the potential loss of jobs, coupled with a transition to a society in which there are a few people who generate considerable value, and a large number of people who simply will not be needed.
"AI is generic technology. We can do with it whatever we want. We can use it to generate better health, education, and transport, to close gaps and inequality, to make the use of taxes more efficient, and to deal with climate issues. And we can also put it to other uses too. When people ask me whether AI is good or bad, I always ask in response: Is the internet something good or bad? The internet is also generic technology. Some 70 or 80 percent of the bits on the internet are dedicated to doing bad and even awful things – pornography, arms and drugs trafficking, smuggling, money laundering, whatever. 20 percent of the bits on the internet; however, are involved in doing useful and positive things."
Q: So the internet is bad then?
"Well, not necessarily. I think that they are one and the same. There is a clear danger that some parts of the population will be adversely affected by AI, as they will not be able to deal with the leap forward. And it is our responsibility as a society to address that issue and help them. That's true. But I don't think that technology by nature is either good or bad."
Q: What is the daily agenda of the person in charge of the national AI program?
Let's begin with the question of why we need a national AI program at all. The answer is: in order to maintain Israel's pole position at the forefront of technology, as it is today, five or ten years down the road. Israel is currently in an excellent position."
Q: This interview is getting better as we go along. We are in a good position?
"In all the relevant metrics and standards, Israel is ranked between fifth and tenth place globally in the field of AI. These standards usually examine how much industry there is and how much capital and human capital there is, and whether or not the regulatory terms and conditions enable growth, and how much academic research there is, and to what extent the government is aiming towards this. I know that newspapers often like to publish headlines stating that 'we have missed the boat' and 'we are in a terrible situation,' but that simply is not true. It is factually incorrect. So, I get up every morning, motivated to act with the clear knowledge that this position is not guaranteed for us. If we do not run at a good pace, we will wake up in two years' time and discover that we have slipped down into fiftieth place."
Q: How do we run at a good pace?
"Firstly, we need to engage in an orderly effort to map things out and check where we are strong and where we can leverage the strengths, and where there are places that need to be reinforced. For example, in Israel, the human capital density is extremely high. The rate of AI experts per thousand people is good. But we are also an economy that is extremely dependent on high-tech and the academic sector, and we are a small society, so we need much higher human capital density. We simply cannot put up with average human capital density. Thus, we invest a great deal of energy in order to ensure that we have academics and additional researchers who are able to build startups. Traditionally, we are good at transferring academic know-how to industry and very good at inter-disciplinary connections. This is also manifested in AI. We are not as proficient at creating shapely and orderly strategic documents, and the same is true for establishing infrastructure that will endure for many years to come."
The Labor Movement
He is 46 years old, married to Gali, the father of three girls, and grew up in Netanya. He served as an ordnance officer rather than a technological unit as you might have thought as per the typical high-tech stereotype. "I wrote lots of software code at high school and I was discharged from the army just before the dotcom bubble burst, so it was fairly easy for me to make my way into the world of high-tech based on my knowledge rather than any professional certificates or diplomas. I have spent most of my adult life in the industry." Most of his professional career has been at Verint, where he served for 15 years in a series of technological management and entrepreneurial positions. His last position in the company was CTO (Chief Technology Officer) of its Cyber Intelligence Division.
At the age of forty, he took leave to write his doctorate dissertation, and then the tender for the position in the Israel Innovation Authority was published; he submitted his candidacy and won it. It is "the most difficult position that I have had in my entire career," he states, "by far."
Q: What have you managed to accomplish in the two and a half years that you have been here?
"Quite a bit actually." One of the greatest obstacles to the integration of AI in Israel is that here we tend to speak mainly Hebrew and Arabic. Anybody who uses NLP (Natural Language Processing) tools knows that on the whole all the work is carried out in English alone." NLP is a branch of AI that enables computers to understand, create, and use human language – the technology used by virtual voice assistants such as Apple's Siri and Amazon's Alexa, as well as chat-bots such as Claude or Chat-GPT.
In order to overcome the language barrier, he says, "We have put in a tremendous amount of work, more than fifty projects that have been supported in translation to date, in databases in which models can be trained, in trained models and models of abstraction, entity, and sentiment extraction models, transcription, questions, and answers, error detection. And we are returning all this work to the community. In other words, this is an asset that can be used by a company or a government ministry or an academic researcher. The fruits are beginning to accumulate and I assume that the majority of them will arrive towards the end of the year. These are projects that return to the community and are used by it. An additional field on which much work has been done is that of regulation and ethics. What is often referred to as Trustworthy AI. How do we ensure that all the apps will guarantee human rights and will be free of error."
Q: And this is something that can be done at the local Israeli level?
"Nobody claims that the Europeans and the Americans will adopt the Israeli law, but there is a genuine intention to work as partners on processes that are currently ongoing globally. We are currently at a unique juncture in time. Nobody in the world is sure of how to address this issue so that we here do have the ability to create a significant degree of impact. Last week, I came back to Israel from negotiations regarding an international charter dealing with human rights, democracy, and the rule of law in AI. We are partners together with a large number of states that have worked together to shape the first charter in the world in this sphere. It involves highly complex negotiations. The Europeans pursue a certain method while the Americans opt for a different approach. The Europeans believe in legislation. Other countries say – we have an existing legal framework, but nothing has proven to be missing, let's begin to move forward and try and understand where things are lacking and then we can fix them as we progress. These are very different philosophical approaches.
"Representatives from some fifty different countries sit in the same room and attempt to reach common ground based on shared values relating to what democracies think about AI. Then they try to squeeze this into one legal text so that everybody can live happily with the outcome. Israel is a dominant player in drafting this charter. There are complete paragraphs in the text that are the result of the work that we have fed into this process. For example, we have made a claim regarding the right to innovation. This is an extremely Israeli position. It states that yes it is true that we do need to safeguard the citizens, but if we fail to create sandboxes or places where we can promote experimentalism and innovation – then all the good that we might be able to reap from AI will just not materialize. The bad too might not occur, we cannot guarantee this, but the good will certainly not materialize."
Q: By the way, has the current regional political situation managed to edge its way into this convention?
"We have been part of this negotiating team for two years now. There were strong differences of opinion as to what the charter should look like, but genuine friendship was forged. From the very beginning of the war, I have been offered nothing other than sincere empathy. People asked how I was and my family and how it is to be here during such difficult times. I must admit, I was concerned. When you travel abroad today as an Israeli representative, there is obvious concern. But I have seen nothing other than genuine empathy. The opposite is true. In the world of AI people will make the effort to come and here what we have to say. I attend forums and conferences – everybody makes the effort to come and listen to us. They ask how you succeeded. What did you do? And we have much to be proud of. For example, in Israel, there are more than 2,200 companies engaged in the generation of genuine AI, not instant AI of the simple 'mix and serve' variety."
Q: What do you call "Instant AI"?
"If somebody uses ChatGPT to provide service to their end customers, that is instant. That's okay, I'm not against it, but he has not conducted any R&D (research and development) in the field of AI. So, we have more than 2,200 Israeli companies engaged in genuine R&D. For the sake of comparison, throughout all of Europe there are more or less the same number of companies.
"When you ask the average person in the street today what is AI, they will usually tend to think of Generative AI – models that generate text and images. But the truth is that this is only the tip of the iceberg for the entire AI spectrum. It is exciting, perhaps even jaw-dropping, but it is still rather small. There are plenty of other AI apps that are changing reality as we know it. For example, imagery processing using AI. It is possible to analyze the results of medical tests or photos of cracks in bridges, to engage in digital agronomy that are completely changing the face of agriculture."
Q: Please explain the issue of cracks in bridges. How can AI help with that?
"I can take a photo of the bridge in visible light or in imagery that penetrates into a much deeper level, and then analyze its structural strength: are there cracks in it? does it need to be renovated or not. Instead of an individual poring over thousands of photos and counting crevices and measuring length – I am not a bridge engineer, I don't know exactly what the standards here are... But I do know that now it is possible to take photos of bridges and send them to a computerized model that will conduct the relevant analysis for me. It will probably do it much faster and more accurately without getting tired.
"We recently had a tour of Kibbutz Yi'ron. There is a drone there that flies over the fields every morning, it photographs and analyzes whether an outbreak of a particular pest has begun in a corner of the field, so that this can then be addressed immediately. Essentially, in terms of global food security, this is groundbreaking technology. There are three Israeli companies considered to be world leaders in the field of digital agronomy. Or for example, examining tissue pathology. Today, an individual sits with a microscope and examines biopsy samples taken from people in order to see if there are cancerous cells. This too is a task that can be performed by AI."
Q: And then you claim that the work of a pathologist who currently sits in the lab is not in danger?
"Today there is a severe lack of pathologists. So, let's now ask just how long a pathologist can sit down without interruption, checking sample after sample, without tiring and not making any errors. What will probably happen is that the number of biopsies that it is possible to analyze at any given moment will multiply by a factor of ten or even a hundred. Only the complex cases in which the computer's recommendation is not absolute will then be forwarded to a human expert. So yes, the profession of a pathologist sitting with a microscope will probably not remain as it is, but there will not necessarily be less pathologists in the world. The manner in which your task is carried out will change. The method of writing code by a software programmer will definitely change. The way in which children are taught at school will change. The way in which you design websites or products will change. The way in which you go over a tax assessment will change. There can be no doubt about that. But that really does not mean that you will become unemployed."
Q: Education – because it is necessary to teach children other things so that they are ready and prepared for this constantly evolving world?
"Also, but the very act of teaching is another issue in itself. In Israel, the average class contains forty children so it is simply not possible to provide each boy and girl with the material they require, tailored for their needs. Those pupils who excel get bored while those who encounter difficulty in understanding the explanations are left behind. It is possible to tailor the content for each child's individual needs: Some children require study that is more visual in nature or involves more audio input or a more conversational approach, some prefer writing and there are those who can run ahead faster or slower. This can apply to a certain field within a subject. For example, if in English the student is proficient at writing but when it comes to talking the language he or she suffers from a lack of confidence, then AI can tailor a study program for that child's individual needs. This will not entail any unemployment in the teaching profession, both in Israel and around the world there is a severe shortage of teachers. Having said that, it is the method of teaching that will change."
Love of the Matrix
Not only does the danger of unemployment raise concerns about the unbridled power of AI, but also sci-fi threats: the machine era in which decisions are made by software in a "closed box" without explaining themselves and without taking into consideration the human factor, progress at such a rapid rate that humans are simply unable to contend with its pace. "The smart robots are taking over the world – everybody and their own favorite movie. I personally am a fan of The Matrix," Katzir states with a smile. "This is a primal fear of technology of any kind. I do understand this but it appears that we have begun to lose it a bit. I don't think that machines are going to take over the world, even though leaps forward such as the one we have witnessed with ChatGPT will continue to occur and at an accelerated pace. Technology is changing and will continue to evolve, and once a year or once every six months we will face some or other technological experience that will take our breath away. Machines will begin to do things that we thought they would never be able to do."
Q: What has stunned you or bowled you over in recent years?
"ChatGPT, although from a purely technological point of view it is not really a jaw-dropping development. One of the winners of the acclaimed Turing Award for inventing deep learning, which forms the technological basis for today's AI, was profoundly insulted. He gave an interview and said: everything had already appeared in my articles, what's new in all this? The answer is that what is new is the fact that it is highly accessible. That all of a sudden, anybody on the street can use this technology with no difficulty at all.
Katzir is concerned about other dangers inherent in AI. "The issue of privacy," he puts forward. "Issues relating to the capacity to explain things. When a machine recommends to somebody to make some or other decision, is that person genuinely able to critically examine that recommendation? There is much still left to do in this field. For example, let's suppose that I develop a system that provides people with recommendations as to their optimum career path. I will take the entire history of Israeli employment from the last twenty years, the individual will then engage in a conversation with the system, and it will say 'you should choose the following profession.' But in Israel, most of those employed in the teaching profession are women and 66 percent of the high-tech workers are men, so if I teach the system using past data and then ask it to forecast the future, it will simply mirror what exists in the data, as it is unable to act differently."
Q: In other words, the system will tell you: you are a young woman – go and be a teacher or a nurse.
"Yes. The society in which we live is full of bias. There is a danger that a decision-making support system will just intensify that bias or discrimination. And generally speaking, when people receive a recommendation from a computer they tend to accept that recommendation. There are numerous experiments in psychology that prove this. But a computer knows how to generate a thousand decisions in a second while a human is able to make one decision in an hour. If we take all the existing biases, then accelerate them by a factor of whatever, and people begin to adopt them without any critical response, we are likely to end up with considerable intensification and reinforcement of the existing biases. This is a process that needs to be addressed, and people are currently working on it."
Q: How?
"First of all, we need to be aware of it. Two weeks ago, we issued an initial call for action to integrate AI in government ministries. Within the call for action, we said: conduct a risk survey and then say how you are going to manage this in the future. In the employment example, after you have built the recommendation system, you need to take the resumes, input them into the system once as a man and then again as a woman, and then see if the system provides the same recommendation or not. After that, take the system six months after it has been launched on the market, and carry out statistical analysis of the recommendations to examine whether there is a bias across one thousand recommendations, two thousand recommendations, 5,000 recommendations. By the way, I do not wish to solve all the biases. If in the name of equality, we will begin to send men for a uterine cancer scan, that would probably be a much less desirable use of taxpayers' money."
As far as the government is concerned, it is not advisable to intervene in every form of bias. "If Spotify makes an error and selects songs that I don't particularly like, then no material damage has been incurred," explains Katzir. "But if there is a bias in the pathology of a biopsy, then that it is clearly much more dangerous. Thus, we conduct risk surveys and then adapt the regulatory burden to the level of the risk. Risk management and proportionality are two principles that enjoy a broad degree of consensus. And it is from here that the split begins. Europe leads the line that advocates the need for broad legislation: in all AI apps whatever they may be, from Spotify to the pathologist, education and taxes, and whoever recommends what articles I should write – in all of these AI is AI is AI, so that legislation will regulate all of them together. In other words, they divide all the AI apps in the world into three parts: unacceptable risk, which are practices that the Europeans are not prepared to entertain under any circumstances, high risk and low risk."
Q: What enters the category of unacceptable?
"For example, biometrics in the public sphere. Facial recognition cameras on the street. Neither are they prepared to see 'social scoring' – a situation in which my civil rights are intertwined with my behavior according to some or other AI tool. Certain disciplines are in danger – for example, systems that are related to safety, such as autonomous vehicles. The implication of this is that anybody seeking to manufacture or market a product in Europe will be required to comply with the regulations.
"In contrast, the US, Canada, Britain, Israel and additional countries say that: examination of the app must be carried out within the specific context. It is not possible to conduct across-the-board examinations. One of my favorite examples from the world of medicine says the following: If I had a medication for onychomycosis (fungal nail infection) that might cause me to go deaf, then it is most unlikely to be approved. However, if I had a drug for a virulent, incurable form of cancer, which too might cause me to go deaf, then I would probably try it out. In other words, the context is of tremendous importance. And the Israeli approach opts for this method: examining the regulatory burden within the context by a regulator who understands the specific area of practice. The regulator must truly understand the potential benefits and the risk involved.
"Regulation is not necessarily legislation. Much of it is interpretation. I am not sure that primary legislation is needed for this. To impersonate somebody and then sell that person's house is already illegal today, there is no need for any new law to state that this is illegal. To expose private databases is already illegal today. This might be done using other means because of AI, but there is no legal vacuum. A great deal of interpretation is needed."
Q: Our strength lies in our chaos
The TELEM program – National Infrastructure Forum for Research and Development – headed by Katzir is also responsible for awarding grants to research students with a view to bolstering the human capital, and also for establishing IT infrastructure, super-computers accessible for young startup companies and for researchers in the academic world. "There is a global lack of IT infrastructure and we need to make them accessible in Israel and at a fair price that will be able to support innovation. It is happening here. We are engaged in activity to encourage experimentalism."
TELEM is a partnership of all the government R&D entities: the Israel Innovation Authority, the Planning and Budgeting Committee (PBC) which represents the academic world, the Directorate of Defense Research and Development (DDR&D), often better known by its Hebrew acronym MAFAT, which represents the defense & security sector, the Ministry of Science, Technology and Space, the Ministry of Finance and the Israel National Digital Agency, whose function is to integrate technology in Israel's public service sector itself. "That is the core," explains Katzir. "There are additional entities that surround this core such as the Ministry of Justice, the Ministry of Foreign Affairs, and so on."
Q: What do you need to fight so that Israel will continue to lead the way in this field? The budget?
"There is a budget. We were allotted a budget of one billion shekels over six years. That is a considerable amount of money and there is a commitment to continue to budget our activity. The key difficulty is in coordinating all the entities and getting them to move forward together. On the one hand, this is something that Israel excels at. We are a small country with a relatively small government. It is much more difficult, for example, for the US federal government to move together."
But they are much more organized.
"That is actually our strength. We are extremely disorganized and disorderly, we always have been, but that is also the very source of our strength. Our ability to set off and proceed with what we called 'the first stage', to begin to run ahead, to gain experience, to see what succeeds and what does not – that is an extremely apt match for our DNA."
Q: In another ten years' time, how will our daily lives improve thanks to AI?
"It will then be everywhere, engulfing each and every detail of our lives, and it will be much more multi-dimensional. Today we have AI tools that are designed to solve a specific problem. In the future, these will be much more versatile and flexible tools capable of connecting input, answers and media of different types. It is difficult to forecast just how this will look in practice. I am convinced that there will be a dramatic revolution in the way in which we conduct research, how we live and how we communicate, as well as the method we use to manufacture products and how we experience the world around us.
"And I am not worried about this change. Technology has always been a neutral player. The good and the bad lie firmly in our own hands. The world has not brought upon itself disaster or destruction with the numerous technological changes that have occurred to date, and I think that this time too it will not wreak destruction. I do believe that on many occasions, individuals do stupid things, but somehow, as a collective group, we usually tend to be okay."
