Towards artificial general intelligence in sport: what will be its impact?

One of the big early developments of 2023 has been the emergence of artificial general intelligence (AGI) into the public consciousness.

 

For those new to the term, AGI refers to a system or software able to display human cognitive abilities, which just like humans, can generalise to new situations it is not familiar with. The learned experience, abstraction and common sense possessed by humans would all be characteristics displayed by AGI in order for it to have been achieved. Put simply, AGI is similar to the representation of AI that we often see depicted in the movies; robots or beings displaying adaptation, learning, creativity and perhaps even sentience. Contrastingly, the AI that we actually have today is none of those things.

It cannot perform any task not assigned to it.

It cannot feel.

It cannot actually even create – it can only use data it has already been provided with. Rather, AI is designed to perform specific tasks, with the focus of either augmenting or replacing humans by improving our effectiveness or efficiency.

 

Probably the most well publicised current examples of AI are Dall-E 2 and ChatGPT. Both developed by the company Open AI, the former uses natural language to rapidly create realistic images or art from text descriptions. The latter is a large language model designed as a chatbot to answer user questions in a conversational manner, and includes such features as admitting its mistakes and rejecting ‘inappropriate’ requests.  

 

I recently asked Chat-GPT to reveal the biggest issues that could potentially influence the future of sport as we currently know it. The system stated climate change, athlete safety, corruption, doping, inclusivity and overuse injuries. Seems reasonable enough and no surprises there. Similarly, I asked it what would be the main catalysts for us reaching new levels of human performance. Advanced physical and mental training techniques, new technology and a better understanding of nutrition and genetics all formed part of its answer. Again, on first read this checks out, although in retrospect I was hardly testing the system’s capabilities, given that a quick Google search might have given me the same information. A number of concerns have been raised around bias in the system, as well its inability to appropriately source its information and perhaps most worryingly of all, apparently simply fabricate occurrences that never took place.

 

I also asked Dall-E 2 (in great detail I might add) to redesign the logo for the upcoming new Season 5 of the podcast I host, One Track Mind. I also uploaded the existing logo and asked it generate variations on this.

The original logo for the One Track Mind podcast.

As you can see, the results are underwhelming to say the least.

Now I know this is far from the main purpose of this tool (and the latter feature is still in beta), but it is just another in a mounting number of examples of just how far we still have to go.

So, with the substantial limitations of existing AI in mind, just how far off then are we from actually achieving AGI? Well, depending on who you listen to, it’s either just around the corner or still quite some way off. To get up to speed on the current state of play, one probably has to look no further than the almost daily banter occurring on social media between prominent AI researcher and author, Gary Marcus and Meta’s chief AI scientist, Yann LeCun. Marcus has continually called into question apparent progress in this space whereas LeCun is considerably more bullish. Only time will tell who is right.

For the record, I am firmly in Marcus’s camp. For those that know me personally, particularly my default sceptical predisposition, that will come as no surprise. But the good news is that, for sport right now, as I outline below, I don’t think this is matters one bit. There are plenty of gains to be made with what we have right now.

 

How will AI impact sport?

 

Despite its many limitations, AI is developing much, much faster than sport. In fact, it’s not even a close-run race. That shouldn’t be taken as a criticism of sport. It’s simply inevitable given the resources being funnelled into AI, as well as the high ceiling on the gains that are there to be made. AI also has a whole lot fewer checks and balances to navigate, as well as a completely different level of risk tolerance.

 

I recently asked people on social media to comment on what opportunities they saw for AI in sport. Comments ranged from more efficient written communication, faster access to required knowledge, and even the design of visual tools to help learning in athletes. Search online for ‘AI and sport’ and you will find ideas and sometimes recommendations for AI’s incursion into just about every element of the sport experience, from new media platforms, to injury prediction and individualised training prescription.

 

Whilst it’s tempting to gravitate towards the latest and greatest developments in AI, doing this in sport right now is largely destined to be nothing more than window dressing, a fad. Sure, sport will always be a competitive environment, with rivalled organisations continually looking for ways to get the edge. But the point is that right now, sports are nowhere near fully realising the far more rudimentary, yet more impactful, benefits made possible from existing AI (in fact that is true in most fields – not just sport).


Take for example something as fundamental as the way in which people structure their work week and the tasks they undertake. We could intentionally make large parts of many current sports roles redundant overnight through automation of low-level tasks. Yet all sorts of reporting, communication, coding and analysis remain as predominantly manual tasks. The opportunity to generate more time, freedom and creativity in our jobs is right in front of our eyes; we just need to take it.

 

Why the lack of gains?

 

I am sure there are organisations out there leveraging AI to make more accurate decisions and improve their efficiency, but for the majority that aren’t the reasons behind the lack of uptake are the same as we see in any other field.  

 

Whilst we may not readily admit it, a lot of the time it’s simply cultural reasons that prevent the opportunities being taken up. The irony of time poorness being one of the biggest complaints I hear and observe from people working in sport is not lost on me. Other times, people are too busy to change their processes (see the classic, if not somewhat overused, photo below), they simply prefer the status quo or in some cases perhaps still just aren’t aware of what is possible.

 

This unwillingness to use available technology for its most obvious and intended purpose is not new. The influential mathematician Merrill Flood talked about the benefits of computing for relief from tasks to provide us more time for creativity and social pursuits back in the 1950’s. I wonder what he’d think now?

 

The path towards AGI

 

Of course, if the day does arrive when AGI eventuates, undoubtedly it would transform almost every industry. But for the moment, AGI is not something that sport needs to concern itself with. When it comes to AI in sport, we need to walk before we can run. And it’s probably just as well. Because the thought, care and due diligence we give to any new method or technology before implementing it with our athletes, also needs to be done with AI.

 

·      What does it mean for an AI using poor data to recommend signing a new player?

·      Who is liable when AI makes a ‘bad’ decision to let an athlete compete and they ultimately become injured?

·      Whose responsibility within a sporting organisation is it to govern and update the AI?


These questions and many more have the potential to dramatically alter how sports jobs look, as well as the skills we need to train for. And they need satisfactory answers.


So let’s make sure that we walk before we run.  

Previous
Previous

Are all sporting decisions complex?

Next
Next

Strengthening Entrepreneurial Ecosystems To Thrive