Dr. Roey Tzezana is a futurist researcher at the Blavatnik Interdisciplinary Cyber Research Center at Tel Aviv University. A specialist in the analysis of innovative technology and its impact on society, in his spare time he works as a magician and a mentalist. He spoke with Israel Hayom on how our lives are being changed by the rapid change in technology and artificial intelligence.
Follow Israel Hayom on Facebook, Twitter, and Instagram
Q: Dr. Roey Tzezana, just a couple of weeks after the official launch of the AI (Artificial Intelligence) engine GPT-3, this would seem to be a good time to understand just what we have been able to learn during this last month. Have we witnessed an earthquake in the world of technology, perhaps across the entire human race, or is this more of a slight, and perhaps not so substantial shift?
"Allow me to start by saying that the launch of ChatGPT was accompanied by a mass PR campaign led by OpenAI, which is estimated to be costing the company several million dollars a month – required in order to operate its entire computing power. This is a chat that is based on a linguistic model, and one of the best such models in the world, which was released on the market about two years ago, but which has only recently become more accessible to the general public, who now understand just how impressive it is. The recent versions of the linguistic search engines are so sophisticated that they touch on every service, software, and even computer game. It is clear that we are currently poised on a springboard, and we are now seeing things that only two years ago we could only dream about."
Now speculations have begun to arise as to the next version of GPT.
"And everybody is guessing. We are pretty certain that it should be able to take into account more parameters than GPT-3, work with longer texts, read them, and draw insights from them. One of the guesses is that GPT-4 will not only refer to text, but also to pictures, video, and music. In other words, it will give its opinion on music by Bach, and possibly also produce pictures, movies, and tunes by itself.
"GPT-3's current text capabilities are still limited, but we may assume that the next engine will be able to take in much greater volumes of text and might even be able to write a complete book in accordance with the style that we request from it. It might not be an amazing book to read but could well be a good starting point. Its ability to write code will also improve, and it should be able to read and write better in any language that is well-documented on the web."
Q: One of the interesting surprises with the current engine was recorded when GPT-3 unintentionally learned Hebrew.
"This completely took us by surprise. Nobody intentionally taught it Hebrew, but this sly fox managed to find sufficient information on the internet and thus succeeded in understanding the language. So now, all of a sudden, it can read and answer in Hebrew, even though it still has a considerable amount to learn. Later on, we assume that it will begin to speak in a more focused manner and to the point, it will create more precise context and speak less 'garbage'. And of course, it will be able to respond to more current questions, real up-to-date issues, as for the moment GPT-3's information is relevant only to 2021."
Q: The New York Times reported that Google declared a state of emergency within the company in response to the launch. Do they really have something to get uptight about?
"Yes. GPT-3 does not refer you to sites containing the answers, but it gives you the answers itself. Google has now come to understand just how much people seek answers without having to engage in tiresome quests in search of them. The good news as far as Google is concerned is that they have even more sophisticated AI capabilities than ChatGPT. But to date, Google has not had any good reason to shoot itself in the foot – as if AI provides us with an answer, then we have no need to engage in a search and consequently, you will not be exposed to adverts, which clearly has an impact on Google's profitability. They were also concerned about an engine that might produce garbage and thus would undermine their credibility."
Q: What about Microsoft? Although reports this week indicated that it is about to dismiss 11 thousand employees, but on the other hand – it has invested huge amounts of funding in OpenAI. Is this its golden opportunity to rule the roost once again?
"I think so, this is a great opportunity to integrate cutting-edge AI into its products, and this is definitely a very interesting prospect."
You have good reason to be concerned
An interesting opportunity from another sphere that is worth commenting on relates to the potential employment-related implications of GPT-3. Do you foresee a scenario in which managers will sack some of their employees as AI will render their work redundant?
"Yes, and this is already starting to occur. Only recently, MIT issued a study according to which a large portion of the historical wage inequality is related to automation processes. Whoever can harness technology and apply it to his work will earn a higher wage, and whoever is unable to do so – is likely to be fired, or if not, suffer a freeze in wage rises. Over the course of time, there will be a decline in the need for people engaged in routine working tasks."
Q: What is occurring is very interesting, as in the past we used to think that blue-collar workers would be the first to be replaced, while in practice, certainly in view of ChatGPT, those who are threatened with being the first to be replaced are actually the white-collar employees – managers, engineers, creative staff. Should we be anxious?
"I warned of this phenomenon over a decade ago. Firstly, we all need to be concerned about our work all the time, especially in this modern age. Until the appearance of AGI (Artificial General Intelligence), which is equivalent to human intellect or possibly surpasses it, it will be difficult to create a computer capable of replacing all the tasks performed by a programmer, but that doesn't mean that it is impossible to replace ten percent of his tasks even today. This would free up a considerable amount of time.
"In any event, senior managers in a variety of industry sectors are beginning to quietly, carefully examine just to what extent the AI engines are capable of carrying out tasks of junior workers, programmers for example. In many cases, they discover that AI is capable of doing the work of these junior employees, and perhaps even replace them."
Q: What is the implication of this?
"Very simple: If a company has to decide whether to hire an additional ten rookie programmers, or to hire eight and fill in the gap with some or other AI engine, it is clear which option they will prefer to choose. And that is only the beginning.
"They might not attempt to replace everybody tomorrow morning, but it will be sufficient if they can complete ten percent of the tasks of a specific professional for the shockwave to occur, and the shockwave is already happening. But it is important to remember that the idea is not to fight technology but to use it and leverage it in order to fulfill ideas and aspirations more efficiently. When the first camera was produced, artists said that this is not art, and it wasn't long before they discovered that actually, it is."
Q: Do you sleep soundly at night?
"Absolutely. In the last year and a half, I have been corresponding with GPT-3 and other engines, and have been asking them to develop future scenarios for me. I often hold my head in my hands and say to myself – Good Lord, I am speaking with a machine, and if it is no cleverer than I am, then it is quite clear where this is leading. Yet, on the flip side, I also say: What an honor it is for me to be able to speak with the 'child' that has been developed by humanity, which will soon be able to help the entire world."
Q: But doesn't that child embody something that actually impedes the passion for creation, the human desire to advance and conquer new peaks? It is to an extent extinguishing the spark of creativity inside us.
"I use it to write songs, and it produces work that is not bad at all. But when I send a song written by the engine to another person, it is rather devoid of emotion, as people know that there is no genuine intention of another human being behind it, which involved a creative effort. You could always compete in a sprint race with Usain Bolt and beat him by riding a motorbike, but I think that we would all much rather to prefer to see human sprinters running themselves against each other. This is also true of music – the most exciting thing about music is the ability to create despite all the difficulties and then to gain a human victory. Thus, even when AI reaches its own peak, it will not be able to compete with the ability to animate and inspire people as a result of the effort invested in producing the final product."
The problem: We are dead
Q: Both in terms of the creative aspect and in general, do you believe that the most important issue is not necessarily ChatGPT3 itself, but rather the implications of its follow-on engines?
"Definitely. In terms of the long-term overview, GPT-3 is another milestone, but no more than that. This is another milestone on the road to attaining AGI and ASI (Artificial Super Intelligence) that will be able to perform every task and do everything better than humans can. As from time immemorial, the most important thing for humanity has always been intelligence, the ability to invent and innovate. Without this, we would never have produced the inventions that have pushed humanity forward to where it is today. The problem is that humans do have a number of issues: we find it difficult to arrive at new inventions and to put them into effect, most of us specialize in only one field, which slows down the process of human development, and in the end, we all die anyway.
"An individual gains experience throughout his entire life, which from a broader historical perspective is only an inconsequential speck, and after several decades everything goes to waste, both as he dies and also as when he grows older he becomes less and less willing to adopt new ideas, and consequently he delays the continued development of innovation. It is also hard for us to share information: so we have a good brain, but it is difficult to convert this into dozens or hundreds of excellent brains that can exchange information with each other."
Q: And AI is capable of circumventing all these problems.
"Yes, as such, it is apparently the greatest revolution ever that humanity will experience. It will enable us to fulfill our desires and needs: more health, prosperity, energy, and technology. AGI will be able to study any subject matter that a human can learn, transfer know-how between various spheres and make integrative conclusions. The significance of this is that any new insight, theory, and technology will serve as fertile ground from which AI will be able to rapidly build new theories and technologies without having to invest years of effort to do so. By the way, it is possible to assert that GPT-3, at least in terms of text, is already AGI."
Q: It is pretty clear from what you say just what AGI will be able to do. Please, explain some more about ASI.
"If Einstein would have lived forever – just try and imagine what insights and breakthroughs this could have resulted in. So, such intelligence could be equivalent to a thousand Einsteins, and not just in terms of physics but in every field of knowledge. This could lead to a tremendous improvement in the quality of life and the general capabilities of all humanity. This is what I am so excited about."
Q: When will we be able to see ASI actually being put to practice here?
"In surveys, we conducted among AI researchers, the Asians are optimistic and believe that we should be able to see it working within 25 years. Western researchers are more pessimistic and tend to cite the year 2060 as the approximate date for this. The famous inventor and futurist, Ray Kurzweil, believes that we will develop AGI more or less by 2045, and I agree with his position."
Human ethics are still here
Q: A critical question for your approach is who will be in control of Artificial Intelligence in the future.
"If it operates on an IaaS (Infrastructure as a Service) cloud platform provided by one company, then clearly it will be subject to highly centralized control. Alternatively, it might run on all publicly owned devices, as is the case today. One of the key dilemmas surrounding this is the agent dilemma. Ostensibly, AI could be the perfect agent, as it has no will of its own, and thus it should represent us extremely well, just as a good agent would. However, we really should take into consideration the fact that somebody has developed this intelligence and continues to invest in it as they want to make money. It thus follows that AI is certainly not bias-free. Their bias will clearly be intended bias that the companies have implanted within the AI. Additional, and potentially more severe bias, would be the existence of unintended bias."
Q: For example, bias against minorities or the dissemination of offensive content.
"Correct. The reason for such biases and for the dissemination of offensive content is that AI studies patterns and structures based on past experience. When AI advises a judge in relation to the chances of an Afro-American of returning to crime compared with a white individual, it will tell him that criminal recidivism is much more widespread among Afro-Americans. You can call this racism, but the truth is that the answer provided is based on historical patterns that the AI has identified."
"But in this manner, AI is simply perpetuating an existing state of affairs and preserving social gaps. We should thus be asking: If the assumptions provided by the AI are correct, then what should we as a society be doing about it? How should we be using this knowledge to help people become rehabilitated more easily, based on the understanding that the current pattern of behavior is not the ideal one?"
Q: How can we ensure that the right people have control over AI and not just small groups that might put it to potentially dangerous use?
"Rulers in totalitarian states are reveling in the thought of what they can do with AI. Vladimir Putin has already said that whoever controls AI will control the world. Dictators are always anxious about the possibility of the public rising up and overthrowing them, and this applies to all states.
"If AI had existed at the time of the Stasi or the KGB, it would have enabled them to penetrate everybody's smartphone, to monitor and oversee conversations or anti-government content. In a similar scenario, that is occurring today in certain countries, if we are to take a photo in a location that the government does not want to see appearing in the media headlines, then the AI used in operating our smartphone will enable only limited sharing of that image, if at all so that hardly anyone will be able to see it. If democratic governments are allowed to operate AI in this manner, we will soon find ourselves in a completely monitored society, which will cause grave damage to democracy, its citizens, and to national resilience.
"Thus, whoever asks me when will Artificial Intelligence conquer the world, I reply that currently there is no reason to worry about AI, but we should be concerned about who is in control of it, as it is already beginning to guide our actions and deeds: it advises on how much tax we should pay and almost determines how much time people should spend in prison. Thus, once again, the real question is who dictates what value system and considerations should be applied by AI. If the answer to this is the state, then it must be subject to extremely tight supervision – governmental, public, and legal."
Subscribe to Israel Hayom's daily newsletter and never miss our top stories
