AI steps out of the screen: The era of intelligent machines begins

Artificial intelligence (AI) is evolving at a rapid pace, transitioning from narrow, specialized algorithms to systems capable of fundamentally transforming the world. This view was expressed on October 10 by NVIDIA Vice President Rev Lebaredian during his keynote at DigiConf. In his presentation, he outlined the progress of AI, focusing in particular on its evolution and the emergence of physical AI.

The Evolution of AI: From Classification to Generative Models

Lebaredian explained that the AI revolution began in 2012, when researchers at the University of Toronto created the neural network AlexNet, which solved the image classification problem. He pointed out that for fifty years, scientists had been unable to design an algorithm capable of distinguishing a dog from a cat — something AlexNet finally made possible. This breakthrough marked a fundamental shift in programming: instead of manually writing code, developers began feeding data to computers, allowing them to generate algorithms on their own.

He noted that the advent of transformer architectures in 2017–2018 ushered in the era of generative AI. Researchers had effectively processed the entire internet through supercomputers, producing models capable of understanding language on a general level. Models such as GPT became universal tools able to perform a wide range of tasks. The release of ChatGPT in 2022 demonstrated the power of these technologies and made AI accessible through a simple user interface.

Physical AI: A New Technological Era

Lebaredian identified physical AI as the next major step in technological progress. He explained that this stage involves taking intelligent digital agents and giving them physical embodiments so they can act in the real world. Examples include self-driving vehicles, factory and warehouse robots, and systems capable of perceiving their surroundings through cameras and making autonomous decisions.

He emphasized that the physical AI market spans industries connected to the material world, collectively valued at around $100 trillion, compared to roughly $5 trillion for the information technology sector. For the first time in history, computational technologies can act as a genuine amplifier and accelerator for these physical industries, rather than being confined to the digital domain. He stated that applying AI to the physical world could accelerate progress across all sectors by automating tasks that previously required human labor.

Supercomputers as the Foundation of AI

Lebaredian compared supercomputers to industrial infrastructure, describing them as essential to AI development. He explained that a supercomputer is not merely a large machine but an integrated system functioning like a factory. NVIDIA, known for its powerful chips, builds entire systems that combine thousands of components to process massive data volumes. By 2019, three of the world’s twenty most powerful supercomputers were operated by NVIDIA, underscoring the company’s leadership in the field.

He stressed that supercomputers serve as fundamental tools for computer science, much like particle accelerators do for physics. Without them, he argued, it would be impossible to push AI forward. These systems convert data and energy into “tokens” — units of digital knowledge that form the basis for new AI models.

AI Factories: The Next Industrial Revolution

Lebaredian compared the current AI revolution to the invention of electricity. He said that new “AI factories” are being built to transform energy and data into intelligence. These facilities process information into knowledge, a valuable product that can be applied in virtually any industry. However, he noted that such factories require substantial resources, including land, energy, specialized infrastructure, and high-speed internet connections.

He distinguished AI factories from traditional data centers, which merely store and transmit information. According to him, AI factories resemble production plants that take raw materials and energy to produce high-value products. This, he argued, opens up entirely new economic opportunities, allowing the creation of value comparable to manufacturing cars from metal and plastic.

The Future of AI and Its Impact

Lebaredian envisions the future of AI in the rise of autonomous systems capable of acting as independent agents. He explained that humanity is transitioning from tools operated by humans to systems capable of making their own decisions. Such systems are already active in digital environments — handling customer service, managing schedules, and optimizing workflows — and will soon expand into the physical world.

He also emphasized that AI can generate economic value independently, functioning as an “artificial workforce.” This capability, he said, is particularly significant for countries with smaller populations, as AI can boost productivity without relying on human labor.