Posted on June 20, 2018 by staff

The AI Briefing: What are England’s World Cup chances?


An artificial intelligence system designed to predict the outcome of the World Cup has concluded that the famous trophy isn’t likely to be coming home to England.

The AI, developed by researchers from German and Belgian universities, factored in each country’s competitive advantage by weighing up its GDP, population, FIFA rankings and bookmaker odds.

Taking these factors into account, the AI then produced 100,000 simulations of how the tournament could go.

If the AI is to be believed, then England will finish second in Group G, with a 7.1 per cent chance of winning the tournament outright. Which is probably greater than the chances most of us give our young side…

The AI predicts that Spain is the team to beat, with a 17.8 per cent chance of winning, shortly ahead of Germany with 17.1 per cent.

Whilst the research is interesting, does anyone really want to know the scores before they happen? (Anyone who remembers Paul the Octopus would say that they do indeed…)

AI takes time off to make all the memes

It would be great if artificial intelligence researchers could focus on solving real-world problems like world-hunger or election-rigging, but even the most ardent AI scientists need a break.

Enter internet memes, the captioned images you love or hate but can’t escape.

Thanks to the sheer number of ‘success kid’ and ‘awkward seal’ memes created over the years, the images have become a large and valuable data-set from which neural networks can gain insight.

In a whitepaper titled ‘Dank Learning’, lead scientists Abel L. Peirson and E. Meltem Tolunay seized that rich source of valuable time-wastery to create a ‘meme-machine’.

Their AI read and analysed 400,000 existing memes before creating its own, and with the help of human testers, rated the ‘hilarity’ of its output, which even fooled some humans in a blind test.

They say comedy is subjective, but I think we can all agree that this is cutting-edge internet tomfoolery.

Facebook creates a real AI-opener

In the past, blinking at the wrong time could ruin a treasured family photo, or give the impression that you miraculously fell asleep in front of the Leaning Tower of Piza.

Now, the AI experts at Facebook have solved the problem of the age-old mid-photo blink.

There are already solutions to this problem, which ‘paint in’ eyes, but existing software uses photos of other people’s eyes as a reference, which often leads to uncanny results.

Facebook’s new approach makes use of its liberal access to all your other photos.

In simple terms, it analyses other photos of your open eyes to create ‘photo-realistic, personalised in-painting result’, or an approximation of how your eyes might look if you’d just managed to keep them open for that split second.

The AI with a voice for radio

As a radio journalist, losing the ability to speak is the stuff of nightmares. Unfortunately for US journalist Jamie Dupree, it became a reality.

A rare neurological condition left Dupree unable to form words or control his tongue, which effectively ended his on-air career.

After turning to the keyboard whilst trying to find a cure, Dupree heard of a new technology which can recreate bespoke voices from existing recordings.

With a huge back-catalogue of broadcasts, and with the help of Edinburgh-based software company Cereproc, artificial intelligence and neural networks have brought his voice back to life.

Wi-fi wall x-rays beg the question: why?

New research from MIT can sense people through walls using only Wi-Fi signals and a bit of AI to patch things together.

The technology analyses the disturbances in Wi-Fi signals caused by bodies, and then a neural network generates a ‘confidence map’ from the data, which makes an estimation of the body behind the wall as a 2D stick-man.

That’s the ‘how’, but if you’re like me, you’re still trying to figure out a ‘why’ that isn’t a little bit creepy.