Developing modules for artificial intelligence involves much more than just working with numbers and algorithms. Behind every successful project, many other aspects come into play, and they were the subject of discussion during the Back to Business: AI edition.
Three keynote speakers brought their expertise and addressed different AI-related topics. In addition to their presence, three AI-start-ups also pitched their ideas.
Subscribe to IO on Telegram!Subscribe!
Like giving away a newborn child
AI developers at ASML never see their models at work in practice. “It is like giving birth to a child, and then immediately giving it away to somebody else that will take care of it,” explains Arnaud Hubaux, product cluster manager at the company.
The manager explained how the challenge for data scientists and engineers is harder at his company. In fact, developers program without having access to data streams. Also, they don’t receive much feedback “unless the model doesn’t work”, Hubaux adds.
Furthermore, developing AI at ASML also has a different kind of impact on applications. “If Facebook or Google AI make a small mistake and don’t recommend the right product, no one notices that. If an AI in the semiconductor industry makes a mistake, it costs you millions”, Hubaux stresses.
That’s why trust is really important. According to the manager, showing the products’ weaknesses from the outset is a good way to establish good relationships with the customers.
AI can serve multiple purposes, as devices nowadays contain sensors that gather an endless amount of information. Consequently, the possibilities at the edge of AI are limitless.
“AI is now being driven to the edge, and many more things that seemed impossible before are now becoming a reality. All the calculations are now taking place in the devices, without sending data to a cloud, and that represents an opportunity for us,” says Clara Otero Pérez, – senior director of System Innovations at NXP.
Computer vision is just one of the areas mentioned, as it can help perform tasks that people get tired of.
Otero Pérez: “Visual inspections of chips is a good example. A person looking at integrated circuits, checking for errors, gets tired after a while. A machine doesn’t, and it can see hundreds of images and they can learn to recognize if a mistake has been made somewhere.”
Pushing AI at the edge
“Our technology is a game-changer when it comes to in-memory computing. It takes a memory cell and modifies it so it can compute inside the cell. This way, you can make millions of computations within one computational cycle”, says Fabrizio Del Maffeo – CEO and co-founder of Axelera AI.
Del Maffeo presented its company’s history, the idea behind it, and the vision for the future. Despite its recent foundation, Axelera welcomed aboard over thirty people and plans to hire more at the beginning of 2022.
Axelera AI aims to produce the first samples next year and provide open access to the software in 2023.
Reimagining patient consultation
Autoscriber – represented by CEO Jacqueline Kazmaier – was the first start-up to give its pitch. It uses AI to improve healthcare consultations between doctors and patients by recording the conversation. The data goes into the electronic health record.
“This will also automate part of the administrative process for doctors, giving them more time to care for patients,” Kazmaier adds.
The team is currently running a pilot project at the Leiden University Medical Center.
Applying AI for good
Can AI solve some of humanity’s greatest challenges? FruitPunch AI believes so and is educating engineers on how to apply AI in the real world, through AI for good and peer-to-peer learning challenges.
“We collect those challenges for NGOs and other organizations from all over the world. Then we define a problem where we can use data to solve it. Our community is already made up of 1500 engineers, and we’ve already organized 20 challenges”, says Sako Arts, co-founder of FruitPunch AI.
In the previous challenges, machine learning models were used to identify poachers or to help doctors diagnose COVID-19.
Read more about FruitPunch here
Benefitting from AI on the edge
Scailable helps companies deploy AI models in a range of different applications. It does so by optimizing the machine learning models for a device.
“We developed the Scailable AI manager, in an effort to tackle the challenges related to AI on the edge”, says Benedicte Lochtenberg, Scailable’s CCO. “This way, you can pick a model from our library and get it work in your device. It is a plug-and-play solution.”
The technology is being used in different applications, such as product quality checks, predictive maintenance, and smart logistics.
Developing AI in the Brainport region
The Innovation Center for AI was launched last spring. Several events and seminars have been held since then, and 12 companies have joined the center.
Industrializing AI is one of the goals of the initiative. “Cooperating with other actors within the Brainport region is essential, and we’d like to focus on the application side of AI,” states Paul van Son, innovation manager of the High Tech Campus.