The Future of Computing
28 MAR 2021
We are at the start of a paradigm shift. In order to enable increasingly complex and smart computer applications, new computer technologies such as quantum and neuromorphic computers are on the rise.
Merck KGaA, Darmstadt, Germany will also play an important role in the development of these technologies to further advance digital living.
AI demands new solutions
We already expect our smartphones and computers to know where our car is parked and to provide us with personalized offers based on our habits. In the future, computers will also ensure that our cars get from A to B autonomously, control production within entire factories and will likely discover treatments for rare diseases. Innovations such as these are based primarily on AI and therefore require continuously increasing amounts of computing power.
Today, nearly all computers are based on the von Neumann architecture, which is particularly suitable for very precise calculations of complex models. The architecture has proved itself over decades and has been continually optimized. To support the increasing demand for more data, technology has to push the limits of what is physically possible, one example is 3D NAND. This technology enabled further miniaturization by expanding the microchip architecture into three-dimensional space.
The von Neumann architecture is still indispensable and will continue to be further developed in the future. But it is reaching its limits in some aspects, especially when it comes to Artificial Intelligence (AI). Computers increasingly have to perform tasks that include pattern recognition and instinctive and situational processing of information, for example. These kinds of AI applications require a huge number of calculations, using very large sets of data, to be performed simultaneously. The von Neumann architecture is not particularly suitable for this, especially as regards efficient energy consumption. Completely new computing technologies are needed to resolve this issue, which is known as the “von Neumann bottleneck”.
Just like the human brain, neural networks based on neuromorphic chips need to be extremely flexible and adapt intuitively to unpredictable environments. They learn from experience by using networks, trained on stored data to solve problems which they have not previously seen. This requires tremendous computational intensity and is possible because neuromorphic chips can simultaneously store and process information, just like the neurons and synapses in the human brain. By contrast, ordinary computers run commands sequentially, constantly moving data packets back and forth from the memory to the processor. Neuromorphic chips are therefore not only faster but are also extremely energy efficient. The steady data movements in conventional logic chips consume a lot of energy.
Enabling computers of the future
However, it will be a while until neuromorphic computers are as smart as humans. With almost 90 billion neurons that are connected by trillions of synapses, a human brain has a computing power of about 4 to 5 petabytes. Current neuromorphic computers have about 100 million artificial neurons – only one thousandth the number of a brain. Nevertheless, the technology has the potential to take artificial intelligence and machine learning to the next level in the coming years.
A comparably disruptive potential is also ascribed to quantum computers. Unlike a von Neumann computer, which performs calculations using classic binary bits with a value of 0 or 1, a quantum computer calculates using so-called qubits, which can process considerably more information. These qubits can become entangled, allowing quantum computers to perform mathematical operations and calculations for highly complex models at an unprecedented speed. However, there are still many challenges to overcome since current qubits are very fragile and the slightest interaction with their surroundings can distort them.
We expect that the von Neumann architecture will not be replaced but will instead continue to coexist alongside quantum and neuromorphic computers as all three technologies have distinct advantages in certain areas. That is why we at Merck KGaA, Darmstadt, Germany are actively involved in both further innovating existing computer architectures and paving the way for exciting future computer technologies. In doing so, we can capitalize on our longstanding experience, profound expertise and broad portfolio in electronic materials for microchip production.
Moreover, we not only offer the cutting-edge materials that are required to produce existing technologies but also provide a powerful analytics infrastructure to develop entirely new materials for the computers of tomorrow. This set-up makes us the ideal partner to further drive innovation in computers.