What's not to love? What could go wrong?
"Premier League to bring in AI-powered camera system to speed up VAR process
The Premier League is poised to unveil a revolutionary new version of the video assistant referee system using semi-automatic offside technology (SAOT) which, it claims, will produce the best officiating setup in the world.
The AI-powered system has been developed by the technology company Genius Sports, which will install 28 computer vision cameras at every Premier League ground. The company is confident it can deliver VAR verdicts within seconds and shave minutes off top-flight matches."
More where this came from:
https://www.theguardian.com/football/article/2024/aug/13/premier-league-to-introduce-revolutionary-camera-system-to-aid-var-process
VAR and AI, a match made in heaven
posted on 14/8/24
comment by Vengeance (U23079)
posted 38 seconds ago
comment by Shaun M - supercalifragilisticOrtaisatrocious (U9955)
posted 15 minutes ago
So human intelligence is inadequate then?
----------------------------------------------------------------------
It's not, when outcomes are the result of objective reasoning. It's not a question of intelligence but more of judgement and interpreting laws accurately, effectively and consistently. When we enter subjectivity that's where outcomes lack the consistency required, that is where AI could eliminate the lack of consistency within a set margin of error every single time.
----------------------------------------------------------------------
There's no subjectivity when it comes to off-side - other than whether a player is active/inactive or obstructing the GK
posted on 14/8/24
comment by Shaun M - supercalifragilisticOrtaisatrocious (U9955)
posted 27 minutes ago
So human intelligence is inadequate then?
----------------------------------------------------------------------
Ja606 is perfect proof of this.
posted on 14/8/24
Devonshire spur
--
Not suggesting it is, it's the active and inactive aspect that's created the biggest confusion for officials, not sure if this was an issue when the concept of "daylight" between the players was used. Anyway, semi auto offside takes care of it now. It will be interesting to see how AI deals with other aspects of law breaches i.e. handballs
posted on 14/8/24
Not sure whether it was real, but I do remember a short-form video of someone finessing and gaslighting an AI bot to return 2+2=5. It must have taken ages for it to "learn" if it did happen.
posted on 14/8/24
I think it’s the natural step forward. We have established refs are basically incompetent. If AI proves to be correct as per the rule books, and it has the potential to according to other fields we can hopefully get rid of refs altogether.
posted on 14/8/24
comment by Vengeance (U23079)
posted 5 hours, 16 minutes ago
Devonshire spur
--
Not suggesting it is, it's the active and inactive aspect that's created the biggest confusion for officials, not sure if this was an issue when the concept of "daylight" between the players was used. Anyway, semi auto offside takes care of it now. It will be interesting to see how AI deals with other aspects of law breaches i.e. handballs
----------------------------------------------------------------------
Im not a VAR fan, because it is only as good as those operating it. How many times have we questioned badly drawn pixelated lines to determine off-sides. Its just seems so flawed to me, using imprecise inputs to achieve a very precise output. Then the time it takes.
I can totally get on board with a rapid, AI based, virtual world produced off-side decision. I already feel like i could trust that, even down to the smallest of decisions....a bit like Hawkeye in tennis when it shows the ball clipping the very outside edge of a line, you just accept it and I feel i will just accept semi-auto offside decisions with the same confidence.
My biggest bugbear is the penalty decisions and the 'high bar' bollox. It is very unsatisfactory and highly subjective....not to mention inconsistent within and across games
posted on 14/8/24
Seems there are a lot of people who think AI is actually reliable.
posted on 14/8/24
comment by it'sonlyagame (U6426)
posted 14 minutes ago
Seems there are a lot of people who think AI is actually reliable.
----------------------------------------------------------------------
As if human beings are reliable and free from error 😏
The latest iteration of ChatGPT, the premium version, is highly reliable. It's not perfect by any means but it will only get better with every iteration. We're only at machine learning stage with general AI, long way off full AI consciousness.
posted on 14/8/24
Sure, it's getting better and better, and very fast, but it isn't as reliable as you seem to believe, at least not going on my own experience researching and using it for professional and academic purposes in the field of linguistics. It still has a considerable way to go.
posted on 14/8/24
comment by it'sonlyagame (U6426)
posted 6 minutes ago
Sure, it's getting better and better, and very fast, but it isn't as reliable as you seem to believe, at least not going on my own experience researching and using it for professional and academic purposes in the field of linguistics. It still has a considerable way to go.
----------------------------------------------------------------------
So, the iteration that can engage in conversations (GPT-4o) is not reliable? Gives you inconsistent results for the same queries / questions ?