Artificial Intelligence represents the cutting edge of innovation today, a place where vital issues, from sustainability to public health, are being tackled. It’s technology that allows us to achieve something that would otherwise be impossible – or incredibly time consuming and resource-heavy – if handled by humans alone.
There are hundreds of impressive AI projects to choose from and narrowing down the list wasn’t easy. However, we have focused here on the projects that have made a significant impact and those that are targeting a particularly pressing issue, such as the challenges that have arisen from the pandemic. We have avoided projects that are hitting the headlines and tried to throw some light on those that aren’t necessarily being spearheaded by large corporations. This is our Top 6.
1. AI helps researchers screen billions of antiviral drugs to find potential COVID-19 treatments
Potential COVID-19 treatments identified in 19 days instead of 10 years.
A team at Vancouver Coastal Health Research Institute used AI software to identify potential antiviral drugs to treat people with COVID-19. Instead of the 10 years the screening process would have taken if handled manually by humans, it took just 19 days. The AI programme processed 40 billion molecular structures in that time to find about 1'000 with the potential to target COVID-19. Data collected by the team will be publicly accessible through Github and the Democratizing Drug Discovery with Deep Docking (D5) platform.
"Basically, we’re now looking to democratize certain components of drug discovery. Previously, you really had to have super-computing facilities to do this type of massive prediction. Now, any grad student from any place in the world can access pretty industrial-size computing power because of this."
– DR. ARTEM CHERKASOV, Senior scientist in University of British Columbia
Published in the Royal Society of Chemistry
Source code of Deep Docking is available on GitHub
2. AI learns the normal ‘pattern of life’ for individual businesses, and spots subtle deviations indicative of a threat
Self-learning AI that enables machines to understand a business in order to autonomously defend it.
Darktrace is a cyber security AI develops an understanding of what is normal behaviour for each user, environment and device within an organisation, and autonomously interrupts malicious behaviour as it emerges, without disruption to regular business operations. It works on the basis of ‘early indicator analysis’ that looks at the breadcrumbs of potential cyber-attacks at several stages before they are attributed to any particular actor and before they escalate into a full-blown crisis.
According to Darktrace’s own research, the information technology (IT) and communications sector was the most targeted industry globally in 2021 – artificial intelligence autonomously interrupted an average of 150'000 threats per week against the sector in 2021.
"There is no magic solution to finding attacks embedded in your software suppliers, so the real challenge for organisations will be to operate while accepting this risk. Getting a sense of what is normal for the software you are trusting will be paramount. AI is perfectly suited for this job, spotting the subtle changes presented by a piece of software that has been compromised will be key to fighting this problem in the future."
– JUSTIN FIER, Director if Cyber Intelligence & Analytics @ DARKTRACE
3. AI helps reimagine recycling
Robots assist humans in sorting through tonnes of trash to detect recyclable materials.
Serious bottlenecks prevent the recycling industry from becoming genuinely effective, supporting sustainability goals and creating a circular economy. AMP Robotics has developed recycling robots that represent the perfect symbiosis of human and AI ability, combined to solve a critical problem. Recycling is not an optimum environment for humans – it tends to be unsanitary and even dangerous and, as a result, the industry faces serious recruitment issues. Recycling robots are the ideal solution, using computer vision to detect recyclables. When it comes to sorting through piles of trash, this is one area where delegation to machines is likely to be broadly welcome – and where automation could be transformative in terms of reducing bottlenecks and meeting key goals.
- Robots can pick up 80 pieces of material per minute versus 40 that humans can achieve. As a result, each robot can handle the work of at least two employees, while freeing those workers to do other jobs at the recycling centre.
- Robots are also more accurate when it comes to sorting waste. This will benefit recycling facilities, which operate on low profit margins, as they may be able to improve the amount of recycled material they sell.
"You have all this material that society produces – plastic bottles, pieces of wood, drywall – and people pay for it, but then somehow it has no value once it’s in the dumpster. Why aren’t we using every part of buffalo?"
– MATANYA HOROWITZ, Founder and CEO in AMP ROBOTICS
4. AI helps to build a per plant platform for regenerative and profitable farming
Per plant intelligence helps to monitor every individual plant and kill weeds without the need for pesticides.
Debate over the use of pesticides has raged for years and part of the problem has been the lack of alternatives when it comes to effective farming methods. Now, three farming robots from the Small Robot Co (Tom, Dick and Harry) have been developed to plant, monitor and treat crops autonomously. Crucially, they can kill weeds without the need for pesticides. Instead of chemicals, the robots are powered by AI models that enable per plant intelligence and decisions which take into account agronomy, soil science and market conditions. This could pave the way for a totally new approach to farming, using technology to move away from pesticides. It could also save a struggling industry:
- Farming has seen a 150% increase in costs in the past 20 years.
- Pesticides may no longer be working as there has been an 84% increase in herbicide resistant weeds.
- These farming robots boost productivity too, delivering an increased yield, as well as minimal chemical usage.
"Robotics offer us a real chance to answer the many questions of modern agriculture in responding to climate change, carbon sequestration, biodiversity and of course soil and food security. The light weight, low impact monitoring robot Tom is now on my farm scanning emerging wheat, for the first time giving me a per plant view of my fields."
– Farmer Weekly Farmer of the Year 2018
5. AI helps to keep beaches clean from cigarette butts and save aquatic organisms
Beach rover robot uses AI to clear up poisonous cigarette butts, even those buried in the sand.
Cigarette filters are full of microplastics and leach more than 30 chemicals when they come into contact with water. As a result, cigarette butts on a beach can be incredibly toxic to aquatic organisms. More than 4.5 trillion cigarette butts end up in the natural environment every year, many of them on beaches, poisoning shoreline animals. This beach rover may help clean all that up.
The BeachBot learns from data – sometimes succeeding, sometimes failing and being corrected by human intervention. With the help of beach visitors – thousands of photos of cigarette were contributed by the public to TechTics via Trove, a transparent data marketplace – the robot learns how to detect cigarette butts and clean coastline.
The BeachBot is a prime example of teamwork between people and bots - this is a human-robot interaction where the public can help make the robots smarter. It is also implementing a transparent active learning approach that means AI could have a huge, and much broader, impact.
"With this transparency, a lot of (Trove contributors) feel like they’re part of a team, that they’re doing it together, that they’re actually helping. It’s important for people to contribute to something lasting."
– CHRISTIAN LIENSBERGER, Lead Principal Program Manager at TROVE
6. AI helps to predict the oxygen needs of hospital COVID-19 patients
Federated learning protects anonymity while making data available for research to create AI tool able to predict COVID-19 oxygen needs.
Addenbrook’s in Cambridge and 20 other hospitals around the world together with NVIDIA have developed a tool to facilitate more effective treatment of COVID-19. In the core of the tool researchers used federated learning – a technique that allowed training an algorithm using multiple decentralised edge devices, so that no data leaves its original location or has to be shared with other hospitals. Thanks to this process researchers were able to draw on maximum available data (covering more than 10'000 patients from Europe, North and South America and Asia) and create a model, called EXAM, able to predict the amount of oxygen a patient would need within 24 hours of arrival in the emergency department. The results demonstrate that the algorithm predicts with a sensitivity of 95% and a specificity of more than 88%.
- This project is one of the largest, most diverse clinical federated learning studies to date.
- The algorithm was trained in just two weeks and can now be used anywhere in the world.
"Usually in AI development, when you create an algorithm on one hospital’s data, it doesn’t work well at any other hospital. By developing the EXAM model using federated learning and objective, multimodal data from different continents, we were able to build a generalizable model that can help frontline physicians worldwide."
– DR ITTAI DAYAN, Author of the Study
