Did the Ai judge prove it can outperform human judges with the 118-112 Usyk scorecard?

[MulaKO]

Then humans don't cut it either?

AI can outperform humans in the things you're mentioning.

When the fugh did humans Cut it

[Roadblock]

Cameras won’t cut it until they have some kind of recognition in the gloves ( sensors ) just like the Hollywood movies
That and only that would make any kind of sense to me

https://www.appclonescript.com/boxin...sensor-gloves/

[Haka]

They guess thats why its so flawed, sensors would not be guessing it would be very precise in terms of being clean connect and how hard was the connect, they are being developed now its pretty obvious the benefits attaching this data to AI, if its just video AI will also guess, and dont forget this is not a dedicated LLM its an agent on an LLM.

How do you think they figure effective aggression if it doesnt have a power part of the equation, the advantages to sensors in the gloves would be eye opening.

You can see from video alone if a shot is good based upon how much the opponent moves his body, how long it takes before he fires back, how much the head snaps back etc... It can be done easily.

[Roadblock]

You can see from video alone if a shot is good based upon how much the opponent moves his body, how long it takes before he fires back, how much the head snaps back etc... It can be done easily.

Think on it a bit longer, I can show you video of guys that get it with bombs and hardly react and others get grazed and hit the floor, so as they say a picture never lies well todays its pretty much all lies, why I asked what cameras were used as aI would need at least 5, every side and above, ideally it would need 360 degrees then it will be really accurate in connects but it will still be estimating power, with sensors in the gloves its 100% accurate not guessing, much easier to program as well because the camera's lock position the sensors measures everything and would take little processing to add up the numbers.

I think you will see AI with sensors by the time its mainstream, right now its more gimmicky and has to guess crucial elements.

[Haka]

Think on it a bit longer, I can show you video of guys that get it with bombs and hardly react and others get grazed and hit the floor, so as they say a picture never lies well todays its pretty much all lies, why I asked what cameras were used as aI would need at least 5, every side and above, ideally it would need 360 degrees then it will be really accurate in connects but it will still be estimating power, with sensors in the gloves its 100% accurate not guessing, much easier to program as well because the camera's lock position the sensors measures everything and would take little processing to add up the numbers.

I think you will see AI with sensors by the time its mainstream, right now its more gimmicky and has to guess crucial elements.

Yes but then you have the opponent covering up with his guard up and chin tucked in, so you can distinguish shots were the opponent is braced. I thought the 118-112 score was the only fair one of the bunch.

[Dr Z]

Although the human judges gave a fair enough score of 116-112 across the board, the Ai judge seemed to go one step further and completely omit Fury's punches which didn't actually hit a scoring target, These were the vast majority of Fury's shots. He was hitting mainly gloves, forearms, elbows, in other words Usyk's guard, but Fury rarely landed clean shots. Seems the Ai took this into consideration, which human judges don't consistently do. So was the Ai score of 118-112 the most accurate, and could this be the beginning of the end for human judges?

117-113 was my score. 7 rounds to 4 with 1 round even.

[MulaKO]

https://www.appclonescript.com/boxin...sensor-gloves/

Mucho gracias

[crimsonfalcon07]

Sample size of 1 is meaningless, and it's not like the human element gets removed. Bias is inherent in the programming and the training, not to mention the possibility of backdoors and other means of corruption. As long as there isn't any real effort at accountability, it's just going to be virtue signaling meant to gull the credulous into thinking it's unbiased and objective, which makes it potentially even more dangerous.

Do you think having these AI judges produce cards absent any context is more worthwhile than requiring a detailed explanation from the judge, human or AI, and having a mechanism for throwing out bad scorecards, no matter the source?

[Roadblock]

Yes but then you have the opponent covering up with his guard up and chin tucked in, so you can distinguish shots were the opponent is braced. I thought the 118-112 score was the only fair one of the bunch.

But thats the whole point its guessing, 100% accuracy is not guessing which is how you want scoring, so what is a little braced maybe it was a fighter bracing for shot that didnt land hard at all, there is so many examples that would fool AI as it fools human judges, and was there 5 cameras or just the telecast which I guarantee the telecast missed landed shots obscured by body angles, I dont believe it can work as intended until sensors are added then its perfect.

The bottom line is because you think its score was right that makes it right, thats flawed as hell lol.

[Haka]

But thats the whole point its guessing, 100% accuracy is not guessing which is how you want scoring, so what is a little braced maybe it was a fighter bracing for shot that didnt land hard at all, there is so many examples that would fool AI as it fools human judges, and was there 5 cameras or just the telecast which I guarantee the telecast missed landed shots obscured by body angles, I dont believe it can work as intended until sensors are added then its perfect.

The bottom line is because you think its score was right that makes it right, thats flawed as hell lol.

You must be joking, here is the full highlight instead of the short one. There can be problems with AI but having it combined with human judges is a good development to double check them.