Monthly Archives: March 2021

Interview with Antonio González, researcher specialised in Computer Architecture and professor at the Universitat Politècnica de Catalunya (UPC)

March 2021

“Cognitive computing will play a key role in mobility issues”

“One of the main challenges in the area of computer systems architecture is to improve their energy efficiency”

“European projects are an excellent channel for collaboration between research groups and SMEs”

Antonio González’s scientific publications have one of the highest impact factors in the world. He has registered a total of 53 patents throughout his research career and has received prizes such as the King Jaime I Award for New Technologies or ICREA Academia, among others. He is the founder and director of the ARCO group (ARchitectures and COmpilers), from which he collaborates with numerous companies in high added value innovation projects.

Do you remember when and where you saw a computer for the first time?

When I came to study at the UPC. I’m from a small town in the Province of Teruel, Valjunquera. I arrived at the university at the start of the 1980s to take a degree in Computer Science, thanks to a brochure I saw when I was taking the pre-university course at Alcañiz Secondary School, which made me change from my plan of studying mathematics. The first computer I saw was in the faculty. It had an Intel 8080 processor, and the first language I used for programming was Pascal.

You founded and direct the ARCO research group. What does the group work on?

The area that our work focuses on is computer architecture, which consists of the design of computing systems at block level, and particularly on how to design microprocessors. The group has 20 members, who are mainly engineers, master’s degree and PhD holders in computer science, specialised in computer architecture. Some are also specialised in the part of circuits, compilers and programming languages. We have senior researchers as well as other trainee researchers.

What kind of relationship do you have with companies?

Important and constant. We have always been clear that in this area of research it is crucial to have a relationship with companies, to better understand the challenges and opportunities for the future, complement our training and attain resources for our research. The companies that are most closely related to our work are designers and manufacturers of microprocessors, which are very large companies due to the enormous cost of designing and manufacturing these systems. We have collaborated with many companies like HP, IBM, STMicroelectronics, Analog Devices. But our main collaborator without a doubt has been the company Intel, world leader in the manufacture of microprocessors, with whom we set up a research centre in Barcelona in 2002 with around 50 researchers.

Has the race to achieve smaller circuits come to an end?

We have been very fortunate, because those who work in the development of new technologies for manufacturing integrated circuits have managed to launch new technology every two years with transistors that occupy half the area, are faster and consume less energy than previous models. This is what is known as Moore’s Law. In the last five years, the curve has begun to flatten and is about to reach a plateau or grow very slowly. However, the technology for manufacturing integrated circuits will continue to improve in other aspects such as 3D integration. Doubtless the role of architecture will gain particular relevance to continue to innovate. One of the main challenges in the area of computer systems architecture will be improvement in energy efficiency.

Is this the type of research that ARCO carries out?

That’s right. We are working on studying architectures specialised in certain types of applications that we believe will be particularly relevant. For example, those that are used in graphic applications, GPUs, which are processors specialised in applications that intensively use graphic animations, such as simulators of systems in movement or games for entertainment, among many others. We also investigate what is called cognitive computing, which are systems specialised in tasks that emulate human cognitive functions. The capacity to adopt this type of technologies is very high. We work so that computers can process voice commands and in general communicate with our ordinary language, see what is around them, understand what is happening and take decisions. This has a direct application, for example in areas of mobility.

How can this cognitive computing be applied to mobility?

Cognitive computing will play a key role in many aspects relating to mobility. One of the main areas of application will be autonomous vehicles. Cognitive computing will not only free us from the task of driving but provide us with safer vehicles (reducing accidents), that are more comfortable (so we can do other tasks), and more pleasant (providing new entertainment services) for our journeys. When these systems are attainable for most of the population, mobility will change dramatically: the way in which vehicles circulate, the role of occupants, the management of traffic in cities… The impact will be enormous.

Are there any other star applications?

Virtual assistants. These will be devices that can analyse and understand the environment in which they are designed to operate and they will have the capacity to carry out the tasks we want them to independently and provide support for others. For example, we could have a device that accompanies us to work meetings, recognises the attendants, sums up the main points of the meeting, the reactions of the participants and manages our agenda depending on what has been discussed and agreed. However, to achieve this we need to make relevant advances in the area of cognitive computing. Another field of great relevance is that of health. Cognitive computing will enable improvement in medical diagnosis, the development of personalised treatments that are more effective and less invasive, and improvement in the treatment of rare diseases.

Some of these applications spark ethical debates. How do you resolve them?

This is a very important aspect, although it is true that it affects application designers more than architecture designers. However, we are aware of the debate on the autonomy of machines that can make decisions and we participate in this debate when required.

Is this type of research only accessible to large companies?

No, SMEs can also access this research, particularly when it is on specialised systems. In this area there is a wide range of possibilities for very specific sectors. For example, surveillance systems, access control and domestic robots, among others, have experienced vertiginous growth and had an enormous impact, particularly in the home environment. These systems require sophisticated cognitive functions. It is very expensive to manufacture chips, but a small company can focus on the design and outsource the manufacture. In addition, we have systems based on FPGAs, which are reprogrammable chips to which you can apply your design, and this is attainable for SMEs.

How can companies undertake innovative projects if they do not have departments for this?

By collaborating with research groups like ours, through bilateral work agreements. European projects are another excellent channel for collaboration between research groups and SMEs. In ARCO we have experience with this in the Horizon Progamme 2020 and its predecessors, which promote collaboration with academic and industrial groups that can exploit the research results to innovate in their products.

The European Research Council has funded your research project CoCoUnit with 2.5 million euros over a five-year period. What does the project consist of?

The aim is to develop new computing systems that have cognitive capacities. These systems will be widely adopted in an endless number of environments such as work activities, health, transport and leisure. What we want is to design optimised systems for this type of functions, taking inspiration from how the human brain works and using much less time and energy than the processors that are used today for the same purpose. The practical applications range from object recognition, translation and voice recognition in real-time and with minimal energy consumption, which is crucial for small devices such as mobile phones. We are inspired by how the brain works because it is a system whose energy efficiency is remarkable and much higher than that of current computers for this type of functions, so that we can design new architectures that are both powerful and efficient. To achieve this, we need to rethink architectures and programming. Instead of using programming paradigms that should describe all the steps to take to solve a problem, we investigate systems that learn from themselves, like the human brain. The great challenge is to achieve that these systems can carry out these cognitive functions with the same or more precision than people, but with energy consumption orders of magnitude below that of current computers.

What other topics are you working on?

Another area that we focus on in particular are graphic processors, specifically the design of the architecture for this type of processors. The current trend, which will increase in the future, is to interact with computers through graphic interfaces, images and audiovisual elements… In this area, we have made relevant contributions, especially in the design of GPUs for mobile systems, in which energy consumption is the main parameter to optimise while maintaining high performance, as the demand for animated graphic applications with greater realism is ever increasing, and we continue to work on how to improve this type of systems.

The article was jointly funded by the European Regional Development Funds (ERDF) of the European Union, as part of the Operational Framework of the ERDF in Catalonia 2014-2020.