How much influence are you happy for AI to have?

[alcoblade]

So given that our old manager was ousted for a perceived lack of embracement of modern ways - particularly the new AI model - how much say should the new manager have?

For example, should management be forced to pick Jefferson Cacares for the under 21's?

How much should AI influence team selection for the first team?

Should AI be used to justify much larger fees for players than the manager would himself sanction?

 

[SwissBlade]

So given that our old manager was ousted for a perceived lack of embracement of modern ways - particularly the new AI model - how much say should the new manager have?

For example, should management be forced to pick Jefferson Cacares for the under 21's?

How much should AI influence team selection for the first team?

Should AI be used to justify much larger fees for players than the manager would himself sanction?

There is certainly a lack of understanding of what is being used, so I’d guess that would be the first point to consider

My simple understanding is that the AI Model as you call it is a data based system to initially scout and potentially recruit players. Using attribute data which is available for every professional player can then allow you to tailor your recruitment based on the attributes you want for your players.

The AI part is programming a system to then find those players

The use of data in games and in preparation for games is not new. I think the first time we were wearing the catapult gps vests was possibly back under Clough, but Adkins was the first to make reference to it and using the data when we played Man U in the fa cup.

Adding, let’s say automation to a process by getting a computer to analyse data that the user requests is perhaps the part which is new and speeds up the process

As for recruitment and fees, I don’t think the manager has had control of the financial budget for many years when it comes to buying players. Generally the manager selects the players he wants and puts it to a committee, it’s then a group decision who gets bought

 

[alcoblade]

There is certainly a lack of understanding of what is being used, so I’d guess that would be the first point to consider

My simple understanding is that the AI Model as you call it is a data based system to initially scout and potentially recruit players. Using attribute data which is available for every professional player can then allow you to tailor your recruitment based on the attributes you want for your players.

The AI part is programming a system to then find those players

The use of data in games and in preparation for games is not new. I think the first time we were wearing the catapult gps vests was possibly back under Clough, but Adkins was the first to make reference to it and using the data when we played Man U in the fa cup.

Adding, let’s say automation to a process by getting a computer to analyse data that the user requests is perhaps the part which is new and speeds up the process

As for recruitment and fees, I don’t think the manager has had control of the financial budget for many years when it comes to buying players. Generally the manager selects the players he wants and puts it to a committee, it’s then a group decision who gets bought

We're using AI way beyond selection of potential transfer targets. I think people will be surprised (some excited, some terrified) by how far it will go beyond that.

Regarding transfers I'm "fairly certain" that the "computer says yes" answer has led us to pay way over the odds for at least one player.

Regarding team selection - it's already influencing matters. The selection of the managerial post has and will determine how far that goes.

 

[Chimpan-A]

We're using AI way beyond selection of potential transfer targets. I think people will be surprised (some excited, some terrified) by how far it will go beyond that.

Regarding transfers I'm "fairly certain" that the "computer says yes" answer has led us to pay way over the odds for at least one player.

Regarding team selection - it's already influencing matters. The selection of the managerial post has and will determine how far that goes.

How are you so sure of this? Genuine question, as I haven’t seen any in depth information from the club on the subject.

 

[dexter101]

A friend of mine told his team at work to spend a week putting everything they do through Ai.

At the end of the week, he got them to make a list of what worked and what they felt didn't work. He then told them to do more of the stuff AI couldn't do. It meant his team were more efficient and made use of their skill set more.

I doubt anyone is sitting there saying "tell me the best starting 11" but the amount of analysis that AI could be processing to give really clear pictures of fitness etc is huge.

From my limited experience with AI, the only thing I can't see it being involved in is the actual kicking of the football

 

[alcoblade]

How are you so sure of this? Genuine question, as I haven’t seen any in depth information from the club on the subject.

Very sure, but for obvious reasons I won't elaborate further. You're free to put as much weight on my initial point as that response deserves.

But given we're going much further down an AI route, and have ousted a manager to bring in someone more aligned with that direction, the overall subject is worthy of discussion.

 

[jaybee]

There is certainly a lack of understanding of what is being used, so I’d guess that would be the first point to consider

My simple understanding is that the AI Model as you call it is a data based system to initially scout and potentially recruit players. Using attribute data which is available for every professional player can then allow you to tailor your recruitment based on the attributes you want for your players.

The AI part is programming a system to then find those players

The use of data in games and in preparation for games is not new. I think the first time we were wearing the catapult gps vests was possibly back under Clough, but Adkins was the first to make reference to it and using the data when we played Man U in the fa cup.

Adding, let’s say automation to a process by getting a computer to analyse data that the user requests is perhaps the part which is new and speeds up the process

As for recruitment and fees, I don’t think the manager has had control of the financial budget for many years when it comes to buying players. Generally the manager selects the players he wants and puts it to a committee, it’s then a group decision who gets bought

Your first part about lack of understanding is key…. We’re not sitting in front of ChatGPT or Claude and asking it who we should sign and how much we should pay.

AI isn’t just a chatbot and has been around in various forms for years it’s just the current in thing thanks to generative AI so what would previously have been known as data driven now gets the label AI. Stats like xG are created through AI models,

AI is a set of tools, used well they can be useful and used poorly they can be the opposite - key is understanding what they’re actually telling you so you can then interpret the output properly and understand where the model might be lacking and the level of risk in its predictions. Data driven approaches have worked well for some other clubs in recent years so no problem from me if we’ve got people understanding it well enough to use it well.

 

[brownblade]

So given that our old manager was ousted for a perceived lack of embracement of modern ways - particularly the new AI model - how much say should the new manager have?

For example, should management be forced to pick Jefferson Cacares for the under 21's?

How much should AI influence team selection for the first team?

Should AI be used to justify much larger fees for players than the manager would himself sanction?

AI is never going to replace humans in this context. Can it provide better data better than humans -probably. More data - obviously. Can it replace a human with Football background and coaching badges? Obviously not.

AI is just computers working stuff out/doing what it's told, very quickly.

I'm not sure why there's such concern about AI being used in football data/statistics... At times in the last 50yrs probably every football club has had someone with only a couple of brain cells making key decisions. Which is worse - computers or idiots? Speaking of which I've heard Tinnion, rumoured to be our Director of Football candidate, is not very bright.

 

[alcoblade]

AI is never going to replace humans in this context. Can it provide better data better than humans -probably. More data - obviously. Can it replace a human with Football background and coaching badges? Obviously not.

AI is just computers working stuff out/doing what it's told, very quickly.

I'm not sure why there's such concern about AI being used in football data/statistics... At times in the last 50yrs probably every football club has had someone with only a couple of brain cells making key decisions. Which is worse - computers or idiots? Speaking of which I've heard Tinnion, rumoured to be our Director of Football candidate, is not very bright.

So do you think, for example, Jefferson Caccares should be chosen based on his ability as seen by the management or by the fact that AI tells them he's a good choice?

We can all agree that we should be using data, and AI, to inform decision-making. The question is by how much?

 

[brownblade]

So do you think, for example, Jefferson Caccares should be chosen based on his ability as seen by the management or by the fact that AI tells them he's a good choice?

We can all agree that we should be using data, and AI, so inform decision-making. The question is by how far?

AI isn't going to put its foot down and insist he plays.

It's not the terminator. It's essentially an app or spreadsheet. Or an app that runs on spreadsheets.

The head coach is in charge. Unless the owners are actually mad. Which they're not.

 

[alcoblade]

AI isn't going to put its foot down and insist he plays.

It's not the terminator. It's essentially an app or spreadsheet. Or an app that runs on spreadsheets.

The head coach is in charge. Unless the owners are actually mad. Which they're not.

It isn't. But the owners may well insist you carry out it's instructions to a level most people would be very surprised.

I'm very worried that the head coach is far from in charge now.

 

[brownblade]

It isn't. But the owners may well insist you carry out it's instructions to a level most people would be very surprised.

AI is probably sleeping with the owners to increase its sway. A textbook case of how to win power and influence.

 

[jaybee]

So do you think, for example, Jefferson Caccares should be chosen based on his ability as seen by the management or by the fact that AI tells them he's a good choice?

We can all agree that we should be using data, and AI, to inform decision-making. The question is by how much?

Good choice for what? It seems to me that whatever system(s) we’re using has identified him as a player that has the potential to be worth more than he currently is so we’ve bought him as an investment - as with any investment there’s a risk but you don’t need to get them all right to benefit from it.

It’s not like the manager comes in in the morning and the computer demands he sticks Jeff up front or he’s out on his ear

 

[Slim Man]

We're using AI way beyond selection of potential transfer targets. I think people will be surprised (some excited, some terrified) by how far it will go beyond that.

Regarding transfers I'm "fairly certain" that the "computer says yes" answer has led us to pay way over the odds for at least one player.

Regarding team selection - it's already influencing matters. The selection of the managerial post has and will determine how far that goes.

Untrue

 

[alcoblade]

It’s not like the manager comes in in the morning and the computer demands he sticks Jeff up front or he’s out on his ear

The computer doesn't demand it, no. That's not to say someone else doesn't.

 

[alcoblade]

Untrue

Which part?

 

[Carlton Blade]

It isn't. But the owners may well insist you carry out it's instructions to a level most people would be very surprised.

I'm very worried that the head coach is far from in charge now.

If you don't give a decision making mechanism, of any kind enough ancillary information or context then it's going to draw flawed conclusions. There's no performance related data that says Cannon was a good signing at the time and selection boas can be programmed into AI respondes as easily as any system. Ask it "with our other signings we will create more chances, is Cannon a player who coukd tale them" and it'll respond in a positive manner. Ask it to find a direct replacement for Tyrese Campbell and it won't.

Regardless of what you've heard, I very much doubt that the routines used to ensure the responses are worthwhile are publicly available, neither is the way the data will be presented to those who have public responsibility for results

 

[alcoblade]

There's no performance related data that says Cannon was a good signing at the time

Sorry fella, what do you mean by this?

 

[Gimli]

 

[Carlton Blade]

Sorry fella, what do you mean by this?

The AI/Cannon overpay. Sorry if you were referring to one of the other lads

 

[alcoblade]

The AI/Cannon overpay. Sorry if you were referring to one of the other lads

No I was referring to him - but to be clear what do you mean by there's no performance data?

 

[Carlton Blade]

No I was referring to him - but to be clear what do you mean by there's no performance data?

There's nothing, through everything I've looked into (other than age and the fact he went for £6m 2 years ago based on academy performances) to indicate that Cannon is a player with the potential to be a £20m+ footballer. There's nothing he's doing at an elite level that indicates the potential for huge growth in output, he's mediocre technically and physically by every metric.

I'm not saying AI assistance didn't say "if you put him in favorable circumstances, he shoots well with both feet" but it wasn't a case of "find us a player" and it came back with Cannon. United, at the push of someone had huge bias towards Cannon to start with.

The price is mental. Imagine carrying the Brewster millstone for 5 years then willingly tying yourself to another one. Again, there's no predictory system that thinks that was smart. That was an emotion based decision

 

[Nick Jansky]

We're using AI way beyond selection of potential transfer targets. I think people will be surprised (some excited, some terrified) by how far it will go beyond that.

Regarding transfers I'm "fairly certain" that the "computer says yes" answer has led us to pay way over the odds for at least one player.

Regarding team selection - it's already influencing matters. The selection of the managerial post has and will determine how far that goes.

At the moment you’re hinting at things you know that others don’t. I think you will only get a proper debate if you expand on what it is you claim to know. Nobody is going to simply take this at face value. Otherwise we don’t know exactly what it is we’re debating.

It feels as if this thread ought to be in Rumour Mill.

 

[nopigfansintown]

So given that our old manager was ousted for a perceived lack of embracement of modern ways - particularly the new AI model - how much say should the new manager have?

For example, should management be forced to pick Jefferson Cacares for the under 21's?

How much should AI influence team selection for the first team?

Should AI be used to justify much larger fees for players than the manager would himself sanction?

United have used data in recruitment for donkeys years. It's a myth that our previous staff didn't use it. What we are being told is that they are bringing in a more sophisticated (and no doubt costlier) version and embracing AI. As far as I know AI has been mentioned in conjunction with other things like injury management and prevention. I'm not sure if it applies to signings?

In a sense nothing may have fundamentally changed: you avail yourself of the more sophisticated data analysis (instead of the old version)and get better results. Do you create an amalgam that includes the basic player profile (age, injury record etc) and intangibles like the player's personality, attitude and background, or is it solely data driven? Can data tell you those things? I think they're important so I hope they are part of the thinking somehow.

The more interesting question is perhaps who makes the final decisions on the signings? We had a transfer committee under the Prince. Targets were voted on and then approaches made. The football staff had a vote but were by no means the only ones. I have no idea under this new ownership, although I note the new manager's academic background and how it crosses over with the data led approach. I presume he's involved somewhere along the line.

 

[alcoblade]

There's nothing, through everything I've looked into (other than age and the fact he went for £6m 2 years ago based on academy performances) to indicate that Cannon is a player with the potential to be a £20m+ footballer. There's nothing he's doing at an elite level that indicates the potential for huge growth in output, he's mediocre technically and physically by every metric.

I'm not saying AI assistance didn't say "if you put him in favorable circumstances, he shoots well with both feet" but it wasn't a case of "find us a player" and it came back with Cannon. United, at the push of someone had huge bias towards Cannon to start with.

The price is mental. Imagine carrying the Brewster millstone for 5 years then willingly tying yourself to another one. Again, there's no predictory system that thinks that was smart. That was an emotion based decision

The price is way beyond what we'd have paid if the AI model wasn't employed.

I'm in total agreement that there's nothing in the obvious statistics to back that up - you don't need AI to tell you that. AI tells you things that aren't obvious. That's why it's used. The question is how much it should sway your judgment. In this case I'm suggesting it did - by a long way.

 

[ben1987]

There’s a very black and white view on this forum about AI and it’s implementation in football.

Most premier league clubs are using AI now. I’d be worried if we were point blank refusing to use it.

They’re not going to solely use AI to sign players. Especially big money players intended for the first team. We might see a few speculative punts for the development team but that’s fine.

They’ll use AI to pick players out of pools of 1,000’s, highlight them to the scouting / management team who will then watch the footage or go and watch them in person before a decision is made.

I work in software development and we’re using AI daily. It’s not replacing humans it’s working alongside them to increase productivity, organisation and standardisation. Similarly to how it’s used in sport.

 

[Chimpan-A]

If you don't give a decision making mechanism, of any kind enough ancillary information or context then it's going to draw flawed conclusions. There's no performance related data that says Cannon was a good signing at the time and selection boas can be programmed into AI respondes as easily as any system. Ask it "with our other signings we will create more chances, is Cannon a player who coukd tale them" and it'll respond in a positive manner. Ask it to find a direct replacement for Tyrese Campbell and it won't.

Regardless of what you've heard, I very much doubt that the routines used to ensure the responses are worthwhile are publicly available, neither is the way the data will be presented to those who have public responsibility for results

I suspect that people who are making AI their business are aware of this kind of thing, as someone mentions above, they weren’t expecting Wilder to be typing into Chat GPT “find me a good signing”.
Plus AI learns as it goes, so each subsequent decision should improve based on the outcome of the last one.

 

[Carlton Blade]

The price is way beyond what we'd have paid if the AI model wasn't employed.

I'm in total agreement that there's nothing in the obvious statistics to back that up - you don't need AI to tell you that. AI tells you things that aren't obvious. That's why it's used. The question is how much it should sway your judgment. In this case I'm suggesting it did - by a long way.

I think AI reaffirmed confirmation bias. If you keep telling it you are going to play to a players specific strengths, it'll overvalue the player.

If you'd told it "we are going to surrender possession and use him as a lone forward" it'd have told you to run a mile. This is the ulissue with using AI. You are hostage to the context you give it

 

[Nick Jansky]

The price is way beyond what we'd have paid if the AI model wasn't employed.

I'm in total agreement that there's nothing in the obvious statistics to back that up - you don't need AI to tell you that. AI tells you things that aren't obvious. That's why it's used. The question is how much it should sway your judgment. In this case I'm suggesting it did - by a long way.

Well only know the answer to that when we eventually sell him, not before.

Otherwise, we can’t take you on face value regarding the degree of AI involvement, and I’m sure you understand that.

 

[alcoblade]

At the moment you’re hinting at things you know that others don’t. I think you will only get a proper debate if you expand on what it is you claim to know. Nobody is going to simply take this at face value. Otherwise we don’t know exactly what it is we’re debating.

It feels as if this thread ought to be in Rumour Mill.

It's going down that path, agreed. I was more intending to have a discussion about how far people would be happy to let AI guide us, as it will be way beyond player discovery. It's difficult to respond to suggestions that "there's no way we'd let AI do this" without leaning in what (I think) I know.

I was quite excited about it before, despite previousky wanting to stick with Wilder as manager. The more I hear, the more I'm falling into the "terrified" camp.