20 November 2024
Results feasibility study on hydrogen network in Brainport region presented
- Energy
- Cooperation
- Scaling up energy innovation and products
- Transitioning to hydrogen
- Grid congestion
GrAI Matter Labs is at the heart of where it's all happening now. Artificial intelligence, but with its own specific edge. "Life-Ready AI," the company calls it. What that is? "We deliver brain-inspired chips that behave the way humans do," the company says on its own website. "With artificial intelligence in devices that lets people help act and react in real time. An AI that personalizes the human experience while maximizing efficiency, saving time, money and vital natural resources."
The Eindhoven, Silicon Valley and Paris-based company wants to build a world where AI supports consumer electronics, robots, drones, autonomous vehicles, augmented reality glasses, surveillance cameras "and much more" to make us more productive.
GrAI Matter Labs' roots go back to 2016, but it was in December 2017 that - thanks to an initial round of funding - the company really came to life. Menno Lindwer, GML's VP of IP & Silicon, explains how the company grew from that point into the global organization it is today. “Our goal is to design and market chips so that everyone can benefit from life-like artificial intelligence. Right now, only a few big tech companies benefit from AI. We see that European governments have recently 'woken up' and started to see that it is strategically important to also have such advanced chip manufacturing in-house. "But Europe is not there yet, notes Lindwer. "It is still often the big foreign tech companies that have the knowledge, technology and market. A major reason is that the technology currently only works well in the cloud or on very expensive chips. Moreover, the use of AI raises major issues in terms of real-time availability, speed, power consumption and privacy." GML's AI chips, on the other hand, are small and consume little power. "Therefore, they can be used in all kinds of devices that people use every day. They can perform complex AI tasks independently of the cloud, allowing everyone to benefit from advances in artificial intelligence. Our chips are very fast, making the use of AI fluid and smooth. And because no data needs to be transferred to the cloud, it can then be done with much less worry about privacy."
There are many industries that will need smooth real-time AI, Lindwer adds. "Right now, GML is focusing mainly on high-volume consumer electronics. For example, we have shown that with AI we can split music into vocal and instrumental, again in real time. For example, you can create a karaoke system that allows you to sing directly to your favorite songs taken from your own Spotify account. But the same technology can also make dialogue in movies more understandable. And we have shown that the technology can be used for video, ultrasound and robotics. "
"The breadth of possible applications is enormous."
GML's roadmap still contains several challenges. For example, performing AI tasks takes a lot of energy. Those AI tasks will only get more complex in the near future. Current improvements in chip technology are too slow to keep up. So other techniques are needed to make the ultimate goal ("Life-Ready AI for all") achievable. Says Lindwer, "We have determined that this goal can only be achieved by taking advantage of all possible aspects of 'Sparsity': the fact that current AI algorithms contain a lot of repetitive, redundant or unnecessary data and operations. GML has made a number of inventions that allow us to effectively filter out redundant data and computations without reducing the quality of the AI. As a result, our chips need to do much less work than more traditional AI chips. And as a result, our chips are smaller, cheaper and consume much less power."
GrAI Matter Labs is located at the High Tech Campus in Eindhoven. This location offers excellent facilities, Lindwer notes. "In addition, we collaborate with a whole range of companies here on campus and in the broader Brainport area. But there are also plenty of competitors, both in terms of product, talent and financing. On the one hand, that is sometimes difficult. But mostly it means we don't have to look far for such things."
This does not mean that everything is perfect, he adds. "Chip design and manufacturing are time-consuming and expensive. It easily takes two years to make a chip, and designing AI chips costs at least 10 million euros. We have found that the investment climate in Europe, the Netherlands and Brainport is unsuitable for start-ups in this industry. There is too little money available. And investors seem to prefer to work with shorter-term propositions. Yes, European governments have now determined that in-house chip manufacturing is of strategic importance. A lot of public money is being put into that, too. But new factories are not enough. There must also be companies, like ours, that design, manufacture and sell those chips in-house."
As a winner in the ninth round of the Gerard & Anton Awards, GrAI Matter Labs is part of an already strong tradition. Which of the eighty previous winners stands out for Menno Lindwer? "That would be Intrinsic-ID then: with them we have enjoyed working. Intrinsic-ID has been around longer and has proven itself. They deliver good products with good documentation."