Spoiler Alert: AI already knows who’ll win the Football World Cup

7 min readDec 6, 2022

Welcome to another edition of Bits&Bytes.

Now, as the days get shorter and colder, do you find your focus starts to wander?

I, for one, notice I spend a lot of time in meetings at this time of year, dedicating most of my working hours to comms.

But this year, I’ve decided to change that. I’ve paused a few recurring meetings, giving my marketing team some breathing space while giving myself the time to tackle some long overdue tasks (which I really needed, not least because I’m hosting a webinar next week!).

As for this month’s Bit&Bytes — I’ve decided to focus on the topic on everyone’s TVs right now: The 2022 FIFA World Cup (with an AI spin, of course).

Then we’ll dive into some business insights and, as ever, end with a section on healthcare.

Ready for some red-hot action? Let’s head to Qatar.

AI has a grip on the Football World Cup

Choosing Qatar to host the Football World Cup has raised a lot of questions (on that note, I suggest you watch “Fifa Uncovered” on Netflix).

There have been well-publicized challenges, but let’s avoid politics here. Instead, we’ll look at AI’s role in Qatar because the tech is helping the host nation overcome all sorts of issues. That said, my oh my, has it caused some controversy.

See what I mean below.

Connected Ball Technology

The Video Assistant Referee (VAR) has been a source of much debate in football. But the introduction of ‘Connected Ball Technology’ could go some way to resolving that. Adidas has embedded a motion sensor inside its official World Cup ball, the ‘Al Rihla,’ which can track the ball’s exact location 500 times per second, ultimately helping the VAR make quicker, more accurate decisions.

So no more waiting minutes for a goal to be ruled out for offside. The ball, which Adidas has tested in several tournaments in the run-up to the main event (including the 2021 FIFA Club), promises lightning-fast decisions, all without affecting its performance.

Source: Adidas

Improvements to the VAR

Now let’s give the video assistant referee a little more attention. The VAR is there to help on-field officials make better decisions, using video replays enhanced by algorithms and data points to make the correct call.

In Qatar, VAR will track the connected ball using cameras mounted to the stadiums’ rooves, with 29 data points monitoring each player’s limbs at a rate of 50 times per second. The data is then fed to match officials in the video operation room, enabling them to tell if a player was offside in seconds.

We first saw the VAR at the 2018 World Cup, but this year’s competition promises a much-improved experience. And who knows, maybe it will herald in a footballing era devoid of questionable offside decisions?

Recent experience suggests controversy will always abound… and there was a degree of excitement in looking to see if the linesman’s flag had gone up 😅

Source: The Washington Post

Stadium security monitoring

There are some 15,000 cameras across all eight stadiums, installed to track fan movement on match days. And at the Lusail stadium, the location of the final, the organizers plan to use facial recognition technology, which has (as it often does) stirred up privacy concerns among fans.

Still, security is paramount at the event. And AI could even help prevent an event like that in Indonesia, where an overcrowded stadium led panic-stricken fans into a crush that left 125 people dead.

The cameras will use algorithms to analyze fan movement, with a command and control team forecasting crowd patterns and preventing possible emergencies ahead of time.

Sources: Al Jazeera, Wired

Look away now if you don’t want to know the winner

Now to the most important question of all: who will win?

Well, the Alan Turing Institute has built an algorithm to predict which team is the most likely victor, basing its output on a similar algorithm used for the 2018 Fantasy Premier League.

It relies on a dataset from GitHub, which has tracked the results of every international soccer match since 1872. And to make the model more accurate, it gives more weight to World Cup matches (and more recent matches, at that).

So.. what’s its prediction? Well, the model gives Brazil a 25% chance at winning, an 18% chance to Belgium, and 15% to Argentina.

Source: Turing

How to help your team overcome their fear of AI

In the Halloween issue of Bits&Bytes, I mentioned that one of the biggest fears surrounding AI is when people don’t understand the need for AI. But the truth is: people always fear the unknown (especially if it could put their livelihood at risk).

Just look at the stats: “AI will displace 7 million jobs between 2017 and 2037 while generating about 7.2 million new ones during the same period. As a result, it will lead to net growth of 200,000 jobs.” (Source: PwC)

Unfortunately, this type of reporting will have a negative impact on your employees, making them believe that AI will soon replace them. And that’s why you have to convince them that this won’t happen any time soon.

What’s worse is that articles show AI as “ready to conquer the world,” saving us from work, chores, activities, everything you can imagine. But the truth is that AI’s current capabilities are still far from that because solutions simply can’t do most of what we humans can.

But you need to show this truth to your team, which you can do by discussing case studies. Such sessions allow people to learn, share their fears, and ask questions (they may even suggest ways to improve current systems).

And if you’re looking for a place to start, why not check our AI for Beginners page, which is designed to teach what AI actually is and how it really works.

You can then explore Explainable AI, a technique that helps people understand why a model makes a specific decision alongside signaling potential errors. At DLabs.AI, we recently applied it to a global EdTech platform.

We used the SHAP package to develop an explainable recommendation engine that matches students with relevant courses. The student support team can now see why its system makes a particular recommendation.

And it’s helping them better understand how their AI model works, ultimately making it more trustworthy.

AI can stop hospitals from closing

In almost every report I read on the challenges faced by healthcare, I see concerns about shrinking budgets.

Look at this Syntellis study, for example. It flags how three-in-four healthcare executives say cost containment would be one of their top priorities in 2022. What’s more, more than half say they’re concerned about inflation. But what’s the solution?

Well, according to Accenture, AI could save up to $150 billion annually by 2026, saving community hospitals from imminent closure. Moreover, AI could reduce administrative overheads, reducing costs further by taking on clinical functions.

If you’d like to learn more about these sorts of challenges and solutions, that’s the focus of my upcoming webinar.

DLabs.AI’s thoughts on the European AI Act

Finally, I’d like to invite you to check The Open Loop report “ Artificial Intelligence Act: A Policy Prototyping Experiment. Operationalizing the Requirements for AI Systems — Part I.”

It presents the initial findings and recommendations of the European #AIAct, which was rolled out in partnership with Estonia’s Ministries of Economic Affairs and Communications and Justice, and Malta’s Digital Innovation Authority (MDIA).

DLabs.AI participated in its creation (we were the only Polish participant), providing feedback on several requirements.

The main goal was to assess the clarity, feasibility, and cost-effectiveness of key provisions in the act, seeing if it could fulfill its goal of promoting trustworthy AI while enabling the uptake of AI in Europe.

The report offers insights into the following areas:

Taxonomy of AI actors

Risk management
Transparency, data quality, and human oversight
Technical documentation
Regulatory sandboxes

Maciej Karpicz, our CTO, made a critical point about the complexity of AI models — and their ability to effectively provide relevant, understandable information to the end user, saying “some models are very complex and it takes time to develop and understand those. I don’t have a clear idea how to pass this knowledge to the end user in digestible format.”

Well, I hope that our input will make the AI Act clearer and more technically feasible.

—

On that positive note, let’s bring this month’s newsletter to a close.

As ever, I hope you enjoyed it — and do let me know if I missed something important or, just as valuable, if you found any section a little boring.

I’m still learning, and I’m grateful for all feedback.