The first AI Olympics

T his is the first AI Olympics. In Paris 2024, AI has permeated refereeing, broadcasting, and creating highlight videos in multiple formats and languages. The “bullet time” multi-camera replay system provides cinematic effects, high-quality replays from multiple angles, slow-motion, freeze-frame views, and immersive experiences at twelve venues where athletes compete in beach volleyball, tennis , judo, and rugby.

AI is being used to review the words of every post on social media and to flag unexpected crowd rushes, skirmishes , and unattended luggage. The Olympic AI Agenda is not built around one static objective. The strategy is designed to be a continuous process that leverages insights to unlock the capabilities of AI to p romote harmony, and deeper digitalis ation, improve sustainability and resilience, and reinforce the positive role of sport in society.

The Olympic AI Agenda focus es on optimis ing the Games and detecting talent early . An AI assemblage using computer vision to analyse video s of athletes’ performances was developed to assist talent scouts in assessing potential Olympic hopefuls. AI is being used to make the Olympic Games more sustainable, through a very sophisticated first-ever data capture and energy management system. Using AI to manage the ethics of information organisms (inforgs) who experience life as (onlife) – (both online and offline, both analogue and digital) and who flourish in informational environments (the infosphere ) can help to identify and mitigate r isks at mass gathering events like the Olympics and the Trinidad Carnival.

Cameras powered by AI will identify and flag possible security threats in real-time , and security personnel will then identi fy and confirm those risks, abnormalities , and next steps. In the Paris suburb of Massy, there are approximately 250 security cameras, which exceed the capacity and capability of officials to observe continuously.

The Olympic AI Agenda is a governance and oversight framework. At the Summer Olympics in Paris, 15,000 athletes, 2,000 officials, 5.3 million visitors, and the citizens and merchants of the Paris region which is about 12.4 million will partake in the games in various ways. These figures don’t include the rest of the world , who will view live streams, create posts, and comment on social media platforms.

The International Olympic Committee (IOC) estimates that people will create more than half a billion posts on social media platforms . Assuming the average post is about ten words, it would take 16 years to read all of them if you dedicate just one second of your time to each post . These posts may have the names of spectators, officials and athletes.

Some people will be cheering competitors. O thers will express sentiments of national pride. And then there will be posts that communicate hatred, coordinated abuse, provocation, and even threats of forcefulness that affect the Olympians’ mental health and performance. So in Paris 2024, AI is reviewing the words of every post to keep athletes and officials safe from cyberbullying and abuse.

Moreover, t he Olympic AI Agenda is a collaborative undertaking . The IOC AI Working Group includes AI pioneers, academics, athletes, technology companies, International Federations (IFs), National Olympic Committees (NOCs), Worldwide Olympic Partners, the International Paralympic Committee (IPC) , and several other stakeholders . All of the initiatives that make up the tapestry of the first AI Olympics are nested inside policy and legislative frameworks which are open to continuous revision and updates.

In 2012 the European Commission planned a research project entitled ‘The Onlife Initiative’ to address how the infosphere was refashioning human reality. In 2024 , the EU is pursuing a human-centric, sustainable digital transformation agenda that shapes new ways to achieve ambitions, freedoms, and rights where no one is left behind, and everyone enjoys freedom, protection, and fairness.  The European Artificial Intelligence Act (AIA) introduces a risk-based regulatory framework for AI systems categorising them into four levels of risk: unacceptable, high, limited, and minimal.

However, Claudio Novelli of the University of Bologna, Mariarosaria Taddeo of the Oxford Internet Institute, Antonio Rotolo of the University of Bologna, Luciano Floridi of the University of Oxford and the University of Bologna , and Federico Casolari from the International Research Centre on European Law at the University of Bologna, published an article in “Digital Society,” (March 2024) titled “AI Risk Assessment: A Scenario-Based, Proportional Methodology for the AI Act . ” The authors underscore the nonexistence of a detailed risk assessment method in the AIA for identifying risks in concrete situations.

They argue that the AIA enshrines a “static view of AI risks.” The four risk categories of AI derive generally from application areas of AI. AI is thus treated as a product akin “ to the EU product safety legislation. ” Hence the AIA misses the proportionality judgement between measures of mitigation and the principles and rights involved.

The authors conclude that the AIA misestimates the magnitude of AI risks to the detriment and severity of consequences on values. This makes the legal framework ineffective, resulting in rules that are either too rigid or too relaxed . They propose a framework using hazard chains, vulnerability profiles , and the contextualization of AI risks by drawing on the framework developed by the Inter governmental Panel on Climate Change to address real-world risk scenarios . As the AIA evolves, so too will the LA28 Olympic AI Agenda.