AI: Insights from a Former CERN Scientist and Cledar’s CEO

“Cledar has been strongly linked to the development of artificial intelligence since its inception. Our team was involved in this development and has a good understanding of it, as we’ve got people on board who have been dealing with neural networks for over 20 years. It was a great opportunity for us to be working on large computing projects for the Large Hadron Collider, where we first used artificial intelligence methods on a larger scale to look for new elementary particles and study the smallest structure of the proton.

With the emergence of artificial intelligence, we now expect many profound changes. In my master’s thesis exactly twenty years ago, in which I dealt with the basics of deep neural network operation, I predicted that artificial intelligence would truly get into our lives when computer processing power reaches the level of exaflops. Then, AI would reach critical mass, and this will allow it to be applied to real-life problems that people face at the level of their perception. That’s what is happening right now. Network architectures and training algorithms have also been progressing rapidly. Transformers have been developed, embedding technology has advanced, and multimodal architectures are being created. All these advancements can be attributed to enormous fast memories and computing power, primarily due to the development of graphics cards.

The possibilities for the progress of artificial intelligence have also been determined by access to data. With the development of cloud technologies, corporations have managed to acquire large datasets that correspond in size to the neural network models. The original purpose of the cloud data storage model was to generate extensive datasets for training artificial intelligence. Of course, here’s where the questions regarding intellectual property and legal complications arise, but these issues are likely to be resolved with the further development of artificial intelligence.

I believe that the breakthrough in neural networks could be compared to the creation of the first integrated circuit by Texas Instruments in 1958, about 10 years after the invention of the transistor. It can be said that since then there has been a great advancement in electronics, which is still evident today in Moore’s Law formulated in 1965 and still accurately applicable.

I think that the development of artificial intelligence will now be intertwined with the development of computing power to some extent. The consequences of this will be enormous, and it would probably be difficult to predict social and economic changes, just as it was with microelectronics. In the 60s, no one expected smartphones, the ubiquitous digitalization of communication and business. The same will be true with AI. I expect radical changes in all areas of business and human interaction. At Cledar, we are carrying out many projects incorporating AI into various areas of life, from medical issues (recognizing tumors, supporting knee ligament operations, recognizing the type of tumor in situ during kidney surgery), through human-machine interaction (voice chats, sales processes), to financial processes where we used AI to simulate the economies of entire countries and cities. It’s probably hard to imagine, but there are places in the world where territorial development is planned based on data and precise simulations of political or investment actions are carried out. AI is very helpful here because it allows for the inclusion of details that people normally lose track of.

It can be almost certainly said that competent, intelligent voice assistants will eventually be developed, as well as virtual secretaries, legal assistants, search engines that will provide specific answers instead of links, and generators of projects and visualizations… With time, protocoling meetings and creating summaries will become a thing of the past. All these things are already being created.

I also expect significant progress in medicine, although it will probably be hindered by doctors for a long time. Perhaps it will occur first in less developed countries, where access to doctors is limited. Certainly, AI-based diagnostics already outperform the results of doctors in many areas. Technically, automation of surgical procedures is possible, although it still requires development, both in robotics and multimodal AI models. But all of this is happening right before our eyes.

Robotics is another area where I’m certain of great progress. Until now, robots were precisely designed devices with defined degrees of freedom, managed by special controllers. It was difficult to control machines designed to resemble humans and animals, which have many redundant muscles that we only learn to use during our life (or rehabilitation). Transformer-based models will probably allow for a breakthrough here as well. They will enable the construction of machines with highly redundant drive systems, in which artificial intelligence will learn how to use available mechanisms for specific tasks. This process will probably take several years, but it seems inevitable to me.

Big changes are coming for law firms, consulting companies, education, and language learning. Thanks to current language models, it is possible to create personalized teachers with access to almost all the world’s knowledge. We are already working on this at Cledar. As a fun fact, we are surprised by the usefulness of the GPT-4 model in analyzing ancient languages.

We also believe that AI will bring great progress to science. Already, there are no scientists who can grasp the entire breadth of scientific knowledge. I worked for 12 years at CERN with some of the most brilliant minds, including numerous Nobel laureates. Even they were overwhelmed by the breadth and interdisciplinarity of knowledge. It may be difficult to imagine, but even in the theory of relativity, there are still some mysteries to this day. Scientists are still calculating Einstein’s equations in different ways and studying their implications. Most likely, there will be theories in the future that are only “understood” or better yet, “modeled” by AI, which humans will simply not fully understand due to their complexity. Perhaps AI will eventually help us tackle chaos theory and other computationally difficult problems. It seems that this advancement will be like the development of science through computational techniques.

Numerical computations have led to great advances in science and technology even in areas that were once described by unsolvable equations using elementary functions or algebraic methods. Although Newtonian mechanics and classical physics are relatively simple, it was only through numerical methods that we could land on the moon, build nuclear reactors, and develop modern telecommunications. In the same way, I believe that AI will bring about great advances.”

Hubert Niewiadomski, Ph.D., Founder & CEO of Cledar