Radical Reads: Securing the Olympics with AI

Leah Morris, Velocity Team

    .   

Image Source: Wired

 

The Beijing 2022 Olympic Winter Games kicked off last week, and once again, cyber espionage threatens to derail the games. Known vulnerabilities at this year’s games include encryption weaknesses identified in the MY2022 app – a required download for every Olympian in Beijing. Many athletes, coaches, journalists, and diplomats are reportedly using “burner” laptops and phones to avoid surveillance. 

In 2008, Beijing Olympic security teams faced 12 million threat alerts a day. Over the course of the 2020 Tokyo Olympics, 450 million attempted cyberattacks were identified. When organizations or events face cyberattacks, AI is on the frontlines of defence, providing advanced malware detection, identifying social engineering, detecting anomalies, preventing DNS data exfiltration, and identifying zero-day exploits. 

In 2021, there was an attempted cyberattack a week before the Tokyo games began. Criminals connected a Raspberry Pi device to a national sporting body’s network to steal sensitive data. Security officials leveraged AI to recognize the malicious activity and interrupt the threat. While AI is increasingly being used to enhance sporting events (even helping to write the Tokyo Olympics theme song), the technology’s most important contribution may be its potential to protect the games from crippling cyber threats.

5 Noteworthy AI and Deep Tech Articles: week of February 7, 2022

1) DeepMind says its new AI coding engine is as good as an average human programmer (The Verge)
AlphaCode is an AI system by Alphabet’s AI Lab, DeepMind, designed to write competent computer code. While still in development, AlphaCode was able to draft code that placed it within the top 54% of human coders in competitive tests on a coding platform. The system generates code using Transformer-based language models and was trained on publicly available code on GitHub and fine-tuned on code from programming competitions. The system still produces some bugs, but AlphaCode may lead to tools that broaden access to programming capabilities. 

2) Federated learning and differential privacy for medical image analysis (Nature)
Preserving privacy and security are central to applications of AI in healthcare settings. To date, there has been a lack of publicly available and diverse datasets due to concerns around privacy. Researchers at the Vector Institute for Artificial Intelligence and Layer 6 (both Vector and Layer 6 were co-founded by Radical Ventures founders Jordan Jacobs and Tomi Poutanen) demonstrate a feasible path for sharing and analyzing medical data using a differentially private federated learning framework for complex medical images.  

3) Researchers build AI that builds AI (Quanta Magazine) 

“Artificial intelligence is largely a numbers game.” Training neural networks requires careful tuning to millions, if not billions, of parameters representing the connections between artificial neurons. This training process can take months. Boris Knyazev of the Vector Institute for Artificial Intelligence and University of Guelph and his colleagues have designed and trained a “hypernetwork” to speed up the training process. Named a graph hypernetwork (GHN), its goal is to find the best deep neural network architecture to solve a given task while eliminating the training process. 

4) Unexplored Antarctic meteorite collection sites revealed through machine learning (Science)
Uncovering meteorites lost in Antarctica’s frozen and hostile landscape could provide researchers with more information about the origins and evolutions of the solar system. To better locate the space rocks, researchers trained an AI system with satellite data and data about blue ice fields in Antarctica. The system pinpointed more than 600 potentially meteorite-rich areas on the continent, with an estimated 300,000 meteorites within the ice sheet surface.

5) Why chefs are turning to artificial intelligence (BBC News)
Referred to as “a piano with 5,000 keys,” an AI model has been trained on thousands of flavour combinations to suss out and create new combinations. Human flavour creations are so rare, “you can’t replace them, you can only enhance them.” A team of human flavourists plays this AI-powered ‘flavour piano’ at the world’s largest privately-owned perfume and taste business, Firmenich.

5 AI Predictions For The Year 2030

Forecasting the near-term future presents its own set of challenges, but extending our foresight to a five-year horizon amplifies these difficulties. This week, we feature excerpts from Radical Partner Rob Toews’ latest article in Forbes…

Read More »

© 2023 Radical Ventures Investments Inc.