Today we introduce a new series to Radical Reads. Inflection Point explores the innovations that prominent founders and researchers are most excited about and how they anticipate AI addressing the greatest challenges facing humanity today.
We welcome Himanshu Gupta, Founder and CEO of Radical Ventures portfolio company ClimateAi, which is pioneering the application of AI to climate impact modeling. ClimateAi’s platform predicts weather and climate from 2 weeks to 10 years in the future for areas as small as a farmer’s field, enabling companies to understand and reduce the impact of climate risks on supply chains. As economies navigate the realities of climate change, the ClimateAi platform brings much-needed resiliency to our fragile global supply chain. We asked Himanshu about his vision for the future of AI technology in his field.
What is the most important development in AI right now?
Himanshu Gupta (HG): AI is solving both the big data and the lack of data problem in understanding climate change.
Terabytes of data from global sensors, weather infrared, satellites, and radars is being fed into neural networks to identify patterns of atmospheric and oceanic conditions. AI surfaces patterns in this data to improve the accuracy and reliability of medium and long term weather forecasting.
Predicting long term climate change-related impacts at a local scale is difficult. Regional climate models are computationally expensive to create and some emerging countries do not have sufficient weather infrastructure to be able to localize the climate models at the city or district level. This is where neural network architectures such as GANs have shown promise. If successful, AI will accelerate the climate resilience capabilities of developing economies.
How is it changing your business outlook?
HG: Supply chain planning decisions fundamentally change when extreme weather can be reliably predicted with sufficient lead times. Developments made possible by AI breakthroughs allow us to provide valuable information to our customers to improve their demand or supply planning.
From lithium supplied by Chile, semiconductor chips from Thailand, and pharmaceutical components in Puerto Rico, emerging markets are the lungs of many global supply chains. High-accuracy regional climate models allow us to serve global supply chains and work more effectively in emerging markets.
If AI could solve one challenge, what should it be?
HG: The biggest question today in climate change for both communities and companies is not whether climate change is real but “how will it impact me, my house, my supply chains and what I can do about it?” AI has the potential to generate resilience and disaster recovery strategies for businesses and communities on a global scale.
5 Noteworthy AI and Deep Tech Articles: week of March 7, 2022
1) Ed Clark, chair of the Vector Institute, explains the opportunities and challenges of an AI driven world (The Toronto Star)
“AI is like electricity… if there was a country that said ‘Electricity was coming out, we’re not going to use it. We don’t like it.’ That country is not going to be very well off today.” Ed Clark is the Chair of the Vector Institute for AI, Partner at Radical Ventures, and former CEO of TD Bank Group.
2) How artificial intelligence can help us figure out how life began (NewScientist – subscription required)
Chemists want to know how inanimate molecules on Earth started joining together and replicating themselves. While there are billions of ways this could have happened, Lee Cronin at the University of Glasgow, UK, employs robots to help investigate. His team has set up machines that combine a selection of simple substances – acids, inorganic minerals, carbon-based molecules – to react randomly. The outcome is analyzed, and then an algorithm helps the robot choose how to proceed. In this way, the robot can hunt through vast swathes of information. This systematic exploration of chemical space could lead to discoveries at the fundamental level of chemistry.
3) Waabi’s Raquel Urtasun on the importance of differentiating your startup (TechCrunch – subscription required)
“It’s been an incredible ride. I have to say there is nothing like building what you believe in with a team that you love to work with. There is nothing that can’t be done.” Raquel Urtasun, CEO and Founder of Radical Ventures portfolio company Waabi and former Chief Scientist of Uber ATG, shares her experience starting an AI-first business. The team recently launched a high-fidelity closed-loop simulator called Waabi World to teach AI how to drive.
4) Yann LeCun on a vision to make AI systems learn and reason like animals and humans (IEEE Spectrum)
Humans have the remarkable ability to generate accurate models of the world using our extensive background knowledge from lived experience despite very minimal interactions with any particular environment. Humans do not need to learn every concept to master a new task. Common sense has been an elusive achievement for AI systems. Recent deep learning disciplines such as self-supervised learning (SSL) seem to have a solid foundation for building AI agents that can develop models of the world relatively inexpensively by reusing existing knowledge. AI luminary Yann LeCun outlined an architecture to build AI agents that can create models of the world in a self-supervised way and then use those models to predict, reason, and plan. LeCun’s architecture is based on six core modules, each computing differentiable objective functions and passing them to the upstream modules. SSL is at the centre of LeCun’s ideas. However, each of the six core modules encompasses many AI research areas.
5) An online cursive handwritten medical words recognition system for busy doctors (Nature)
Doctors’ poor handwriting has become a well-known joke but has caused severe problems for pharmacists. Machine learning installed in a smartpen may help busy doctors, especially in high-volume patient settings with minimal digitization. The proposed handwritten recognition technology digitizes handwritten prescriptions in real-time. This study focuses on physicians in Bangladesh who, in some cases, must spend less than a minute on each consultation due to the doctor-to-population ratio: the recommended ratio by the World Health Organization (WHO) is 1:1000, whereas Bangladesh’s ratio is only 0.304:1000.
Share this edition of Radical Reads