Preface
This was written as a draft for my blog post in HCI Advent Calendar 2021.
- Science and analytics in sport is starting to benefit real-world applications
- It is difficult to communicate the feedback to the players effectively
- A short summary of cutting edge research in sports HCI
Table of Contents
AI in Sport
Professional Sports Analytics
Many success stories have been well documented in mainstream publications such as āThe Numbers Gameā (Anderson and David, 2013), āBasketball on Paperā (Oliver, 2020) and perhaps most well known, āMoneyballā (Lewis, 2004). As a result, a growing number of sports teams now adopt specialist roles for analytics.
A Japanese unicorn, started working on tools for professional sports analysis as well as other companies such as HUDL, BePro, StatsBomb etc.
Sports as a testbest for AI Research
Since many challenges arise from sports for numerous research fields, there have been a lot of cutting edge research as well. One of my favourites is the on CVPRā18 paper, āSoccer on your tabletopā.
Earlier this year, Deep Mind published a 48-page paper in collaboration with Liverpool Football Club (LFC) in the Journal of Artifical Intelligence Research (JAIR). It was titled āGame Plan: What AI can do for Football, and What Football can do for AIā.
In this paper, the authors focus on football and explain how the field can benefit AI research in the form of an automated video-assistant coach (AVAC) system. Three fields, Game Theory, Statistical Learning, and Computer Vision are highlighted as being key components in advancing the state of football analytics.
I compiled a list of interesting so check it out,
Remote workouts and Digital Fitness
The demand for home workout kits and remote coaching tools have also seen a sharp rise since the Covid-19 outbreak. The athletic apparel brand Lululemon made a $500 million dollar acquisition for Mirror,Ā a New York-based startup that sells interactive smart-mirrors which can stream live and on-demand workout classes.
Communication = The challenge for AI in sport
Sports Analytics / Remote workout sessions donāt use AI (yet)
No professional sports team relies on AI systems to make any serious decisions. Each team will have a specialist who crunches the numbers, generates the necessary metrics and then translates that into information digestible by the real decision makers, the players and managers.
Lululemonās mirror isnāt exactly an AI mirror either.
Despite previous estimated sales would reach $250 ~ $275 million, Lululemon decide to half those estimates into half, between $125 million and $130 million, just last week.
However, If we were to truly have a fully automated video-assistant coach system, we are missing one important element. I think that missing component is likely to be the interaction between man and machine.
Moravecās Paradox
One thing that an AI coach canāt do that a human coach can is effectively coach the trainee through demonstration, queuing, and assistance. This observation that reasoning requires very little computation, but sensorimotor and perception skills require enormous computational resources is known as Moravecās paradox.
It is comparatively easy to make computers exhibit adult level performance on intelligence tests or playing checkers, and difficult or impossible to give them the skills of a one-year-old when it comes to perception and mobility. (source)
With that in mind, the difficulty in AI coaches likely lies in the difficulty to understand the biological basis of human locomotion.
Current Work
AI FIT
Plug of what I do
kick pro?
Hi, I'm currently researching individual & group human motion at the University of Tsukubaš¤øĀ My interests intersect sports science, computer vision and agent based simulations. See the pages below for more information!
"What I do" in 30s, 5mins, and 10minsPublicationsš¬š§English Posts