Artificial Intelligence (AI) and robotics are playing key roles in the evolution of the manufacturing industry. They are bringing advancement in manufacturing by minimising human workforce, improving efficiency, and even simplifying the whole manufacturing process. This has led to fulfilling increased consumer demand, reduced labour cost, faster decision making, and customised product development.
This article is a recap of the Talkshow OnAir about the implementations of artificial intelligence within organisations. You can watch the whole episode at On-Air.ai.
Today, AI and robotics play a key role in the manufacturing industry by aiding with detection and solving a fault line, automatic control of a system, and demand-based production. While some of the leading companies have adopted AI and robotics in their manufacturing, there are a number of companies that are still playing catch up to their established peers.
In the episode 4 of OnAir, Remy Gieling, founder of ai.nl speaks with Arnaud Hubaux, Product Cluster Manager at ASML and Joris Sijs, Senior Scientist at TNO, who joined the stage with Spot, an agile mobile robot developed by Boston Dynamics. Gieling asks Hubaux and Sijs about their interest in AI, impact of AI in the society, and offers helpful tips for those looking to implement AI in the field of manufacturing, industries and robotics.
An early interest in AI
Our guests had interest in AI as kids with Hubaux calling it software and Sijs calling it robotics. As Product Cluster Manager at ASML, Arnaud Hubaux delivers applications that detect, predict and explain issues with ASML’s scanners. The scanner made by ASML is the lithography machine that turns chip designs into physical chips on silicon wafers. With semiconductor in short supply, these lithography machines are hard to comeby and are hot commodities.
Hubaux says ASML uses AI to “improve the output of these machines and make sure that the uptime is as high as possible.” For him, personally, it is AI’s ability to scale mathematical computing that attracts him to the subject. However, he is a firm believer that “AI will not be a way to create a digital twin of the human brain.” At ASML, Hubaux says they are using AI for image level inference of chip design and at sensor level inferencing to detect defects before they go into production.
Joris Sijs, Senior Scientist at TNO, describes Johnny 5 as his favourite robot before making a quick distinction that it was robot mechanics and not AI-based. He was interested in robotics as a child but saw the ability to bring AI into robotics when neuromorphic, a new form of chip architecture where memory and processing are not separate, became available. Sijs says that this neuromorphic architecture allows AI chips to be put in smaller enclosures without consuming a lot of energy.
TNO, on the other hand, is all about bringing academic research into real life practice. With a big programme on AI, TNO focuses on offline databases, autonomous systems and embodied AI (EAI). The term embodied AI refers to AI designed for virtual robotics and Sijs says that the Spot robot from Boston Dynamics plays a role in this development.
While the Spot robot is being operated by a person behind the scenes, Sijs says the future is one where these robots are autonomous and are able to operate in any open environment. One of the examples, Sijs gives, is that of robots like Spot helping build cars on a big processing chain. TNO is also working with the police in the Netherlands to do inspections of drug labs and study whether they are in compliance with the chemicals they are allowed to store.
Small use cases with big impact
Robotics always starts with smaller use cases but often with much bigger impact. Even with the robotics being tested and implemented by TNO, the use case might seem small but the impact is much bigger. As Hubaux explains, these robots could be used to pick up weeds in the field without having to spray chemicals. At ai.nl, we looked at a use case of AI being used by the City of Amsterdam to measure accessibility.
This is possible thanks to the use of computer vision and machine learning. Speaking of the use case, Sijs also dismisses the myth that AI and robotics will displace human jobs. He says humans will essentially find some other work, which is in line with all the recent findings in the field of AI. Wherever AI is being implemented, automation and smart technology is leading to humans becoming more productive and doing more meaningful work while AI takes care of repetitive tasks.
Arnaud Hubaux adds that AI, even today, struggles to really understand what’s happening at a manufacturing plant. This happens even when an AI model has a 99 per cent accuracy. Hubaux says that AI will still make mistakes that humans would not in the first place. He further adds that AI is too unpredictable in manufacturing right now, which further highlights the fact that AI will continue to augment humans and not supplement them.
AI and Robotics Use Case
Autonomous Shipping: Captain AI, a Rotterdam-based startup, is trying to create autonomous shipping a reality. They are using deep learning techniques to detect where ships are and predict where they are going with the goal of making ships sail autonomously. Captain AI is working with a terabyte of data collected from radar and AIS and run them through a proprietary algorithm to learn the behaviour of ships.
Sijs says autonomous shipping is not that far away as a widely adopted technology. Since it is a traditional domain, the design of autonomous shipping becomes difficult but is beneficial for the entire chain of logistics. Sijs also says building autonomous vessels would be much easier than autonomous vehicles because of all the space available and their trajectory. For autonomous ships, Sijs says that the easy use case would be navigating the open sea while difficulty would be to navigate the harbour.
Hubaux notes that more than the quantity of data, AI will evolve to become all about quality of data. This resonates with a recent statement from adjunct Stanford professor Andrew Ng, who called for smaller data sets with emphasis on expert knowledge in the field over big data available in the lab.
Autonomous machines for hospitals: The OnAir episode also saw CEO and founder of Loop Robots Per Slycke join Gieling in the studio. Lopp Robots is building autonomous disinfection robots for use in hospitals with the goal of reducing the number of infections or transmission. With 5 million deaths so far from COVID-19 pandemic, Loop Robots is building robots that will lower transmission of infectious microbes from touch surfaces.
Slycke says vulnerable people are most likely to contract infections from various touch surfaces, which accounts for up to 40 per cent of infections in hospitals. He adds that humans are really good at cleaning surfaces with visible stains but pathogens are not visible to the human eye. As a result, he says cleaning staff are asked to do work that they cannot see and their managers cannot measure in terms of results.
Loop Robots says that this disinfection activity is something that can be considered repetitive in nature, and requires humans to handle harmful chemicals, disinfectants, which is not healthy for humans. He thus proposes robots as an alternative to carry out this repetitive task without any harm and also leave a digital audit trail.
For its robots, Loop Robots uses existing lamps from Signify but the robot is capable of scanning the room in 3D and determining its own path to optimise and deliver the right amount of light needed for all the different surfaces. Slycke says the biggest challenge for Loop Robots is implementing these robots with the cleaning personnel in hospitals and not technology related.
Hubaux sees these kinds of robots being used in offices and supermarkets, but Slycke does not see a need for such robots in commercial places. He says that these pathogens only infect vulnerable people who could be visiting hospitals or being treated there. However, healthy individuals in office settings or in supermarkets may have immunity against such pathogens. “We as humans need to reduce the amount of chemicals and the medication that we use in an unnecessary way,” Slycke explains.
Future of AI
Joris Sijs believes that the future of AI is one that combines the data driven AI with the knowledge driven AI. He says that by reviving the old fashioned knowledge driven AI, the systems will have much less uncertainties that are common with today’s AI systems that are data driven.
Arnaud Hubaux, on the other hand, sees the future where AI will be much more regulated to focus on the ethical aspect of the technology. Slycke sees a need to get off the hype train and focus on the basics since AI is a very powerful technology and he feels that “emphasis should be on doing useful labour.”