I believe the fuss over AI to be, in many cases, deeply silly. Almost as bad as Y2K.
To be honest, I stopped reading when the author suggested that the costing of contracts could be done by AI. I know finance people who cannot do it, much less AI.
A quick example - what is the cost of going from a hiring model under the CBA where you are required to hire the most senior person out of a hiring hall, vs hiring from any source.
AI is going to tell you that it is a cost to the employer, as there are additional costs involved in hiring from any source instead of being told who to hire.
Even if AI were to understand the benefits of hiring the most qualified individuals - what is the dollar value of the change? Tell me that, brain in a box. It can’t.
Ryan, You are right about AI giving generic answers *right now*...and, often, those answers are either incorrect or partially incorrect.
When I first started investigating AI, asking ChatGPT to write a news article, it was well written, but wrong. That was version 4...and only a couple of months ago. ChatGPT is about to launch version 6 (if it hasn't done so already).
The premise as to how AI could be applied made in the above post are still several years away. [At least I think they are.] 🤷🏽♂️
However, while I don't understand the technology behind it, the more I've studied (and still am), the more I see the potential for it to be applied to huge range of applications.
In May, when I had economist Jon Morrow on the podcast, he laid out a bunch of (rather dystopian) applications that, frankly, I was not convinced would come about. Yet, within two weeks, I began seeing articles confirming his predictions. Namely (just as an example), JP Morgan is investing in AI as a tool for investment strategy/advice (link below). Will the impact be immediate? No.
In the above post, and assuming AI can do what people say it will eventually be able to do, as AI is a tool that is based upon inputs, things like the costing of contracts, determining whether just cause discipline is used in an employee's disciplinary actions, or even "the cost of going from a hiring model under the CBA where you are required to hire the most senior person out of a hiring hall, vs hiring from any source" should not be difficult, provided the data is captured within AI.
On the costing out of a CBA, I am going back three decades to a model we used in my labor management degree program. It was simplistic back then and, presumably, it has been updated since then. With AI, I would think it could even be more advanced.
Thanks again for your input. As I said above, I do hope you're right.
I do think that AI will be impactful in some areas. I would say that it would be highly impactful in investment advice, except that investment banks have already invested heavily in creating algorithms to quickly respond to investment choices (e.g. buying millions in Yen because they are momentarily miss priced vs. the Euro) to the point that I read that banks were putting the servers physically close to Wall Street, because the speed of light is a factor in beating out other banks. Also, I invest only in index funds, because I don’t know anything about investing. So I have no idea how much further it will go.
Rather than the folks who are being paid to talk, and to write things to scare us so that we will watch ads or buy subscriptions, I would look to the people actually designing AI architecture. Also, maybe google ‘Shoggoth and AI’ for how they view the future of AI.
That being said, AI could easily figure out the cost of a 3% increase over last year’s wages, assuming all other factors remained the same. And it could easily factor in payroll taxes and other factors which are generally too much work and have too little impact for me to bother including. But here are some of the factors that will be more challenging for AI:
Hours multiplied by wages - will hours go up, down or stay the same over a three or five year CBA? Even if AI is better at predicting the future than I am, I would trust a conversation with the GM over anything an AI could scrounge up. Is it likely that the State, City or local jurisdiction (i.e. airport, or hospitality sector specific are two I deal with now) will pass a higher minimum wage, which could cause salary compression. Will an AI understand how salary compression can cause you to lose valuable supervisors and managers? Can AI quickly do an accurate salary survey for your area and industry to see if you are already underwater in terms of wages vs. your competition? Can an AI explain (or would even think to do so) the difference between a 3% increase when your wages are already at minimum wage or will be, to help corporate understand that 1% of your increase wage going to happen regardless because of new wage legislation so this could be considered closer to a 2% increase? One challenge many employers have had over the past few years is costing out CBAs based on COVID-times hours vs. post-COVID hours. Some of my spreadsheets use 2019 hours because they are the closest to what we can expect going forward. How would you teach an AI to not use last year’s hours, and to use some other year’s hours instead?
Benefits - I will assume that AI can better predict what the benefit fund increases we will see over the next 5 years than I could. But could AI explain that this proposal has the lowest overall cost, but includes a defined benefit pension plan which could result in millions in withdrawal liability in the future? How would it value that? And how would that change where your company was providing a service, and you could be outbid and lose the contract (and trigger w/d liability) vs. a contract that you control and cannot lose?
Could AI understand the difference - going back to the example above where you are going to ‘any source’ hiring - between a union like the IBEW, who are dispatching trained journeymen, and another who had no formal training program? How would that affect the cost of the CBA?
In other words, there are a lot of subjective factors that go into some, but definitely not all, CBAs.
Last - if the AI told you the CBA was going to cost $1M, but you had no way to ‘look under the hood’ and see what its inputs, assumptions or formulas used - would you trust it? Despite over 20 years of experience, I have yet to cost out a CBA where a more senior exec (or more likely, a finance person at their direction) did not insist on looking through my spreadsheet to see how I came up with my increases over time. Which is as it should be - I’m not a finance person. But you can’t check an AI. It will tell you it’s correct, but it will lie to you - just ask that lawyer who directed an AI to write a brief who learned that everything was made up from whole cloth - even though it swore the cases were real. AIs lie too, just like people.
All of your points are excellent. Broken down, I think there are 'yes' answers to some, and 'no' to others.
I think (as I understand the technology [which is minimal, BTW]) is that it will be wholly dependent on the data input...the more data, the more 'knowledgable' AI will be and more likely accurate as well (in reference to the lawyer, for example).
As an aside, I am having the same economist who was on in May back on the podcast next week. He, by the way, stated several times that he hopes he is wrong.
Also, as a separate aside, I went into studying AI as a potential solution to labor shortage issues and population pyramids. It was the conversation with him that led me to go deeper and conclude this could be bad. (I am also trying to book a professor from UCLA on as well.)
If you're open to it, we should have a conversation offline. I am totally open to the idea that AI is a nothingburger.
the idea that AI could interpret contract provisions or aid in disciplinary hearings strikes me as contrary to the idea that AI frees up people to work on "more complex" matters. Judges and arbitrators routinely disagree on the meaning of contract provisions, so how would AI solve this issue. Further, discipline is fact specific, involves witnesses, etc. AI could not possibly handle this.
To your point in the reply to Tor, as well as this one. AI (more than ChatGPT) "seems to simply compile a bunch of internet searches and put them in a cohesive summary" is fairly accurate.
[When I first started learning about AI, my analogy was 'Google on steroids.' However, I was informed that was incomplete. It is more and, in the future, will be even more than that.]
To that end, imagine a program that can compile all of the arbitration decisions published by the AAA over the last 100 years on a specific topic, as well as the Seven Principles of Just Cause. Then, imagine a program where a termination is input, along with the grievant's argument--even to the point of witness statements. As a former union rep (who's handled hundreds of grievances), if AI is able to grow to where "they" say it's going, I don't know that it is too far fetched to think it can be a useful tool.
I may be wrong...and, I do hope that I am. However, time will tell.
I believe the fuss over AI to be, in many cases, deeply silly. Almost as bad as Y2K.
To be honest, I stopped reading when the author suggested that the costing of contracts could be done by AI. I know finance people who cannot do it, much less AI.
A quick example - what is the cost of going from a hiring model under the CBA where you are required to hire the most senior person out of a hiring hall, vs hiring from any source.
AI is going to tell you that it is a cost to the employer, as there are additional costs involved in hiring from any source instead of being told who to hire.
Even if AI were to understand the benefits of hiring the most qualified individuals - what is the dollar value of the change? Tell me that, brain in a box. It can’t.
well said. AI gives generic answers. ChatGPT seems to simply compile a bunch of internet searches and put them in a cohesive summary (at best).
Ryan, You are right about AI giving generic answers *right now*...and, often, those answers are either incorrect or partially incorrect.
When I first started investigating AI, asking ChatGPT to write a news article, it was well written, but wrong. That was version 4...and only a couple of months ago. ChatGPT is about to launch version 6 (if it hasn't done so already).
The premise as to how AI could be applied made in the above post are still several years away. [At least I think they are.] 🤷🏽♂️
Tor,
Thanks for your input.
You may be right. Hopefully, you are.
However, while I don't understand the technology behind it, the more I've studied (and still am), the more I see the potential for it to be applied to huge range of applications.
In May, when I had economist Jon Morrow on the podcast, he laid out a bunch of (rather dystopian) applications that, frankly, I was not convinced would come about. Yet, within two weeks, I began seeing articles confirming his predictions. Namely (just as an example), JP Morgan is investing in AI as a tool for investment strategy/advice (link below). Will the impact be immediate? No.
In the above post, and assuming AI can do what people say it will eventually be able to do, as AI is a tool that is based upon inputs, things like the costing of contracts, determining whether just cause discipline is used in an employee's disciplinary actions, or even "the cost of going from a hiring model under the CBA where you are required to hire the most senior person out of a hiring hall, vs hiring from any source" should not be difficult, provided the data is captured within AI.
On the costing out of a CBA, I am going back three decades to a model we used in my labor management degree program. It was simplistic back then and, presumably, it has been updated since then. With AI, I would think it could even be more advanced.
Thanks again for your input. As I said above, I do hope you're right.
[Here's the link referenced above: https://www.cnbc.com/2023/05/25/jpmorgan-develops-ai-investment-advisor.html ]
I do think that AI will be impactful in some areas. I would say that it would be highly impactful in investment advice, except that investment banks have already invested heavily in creating algorithms to quickly respond to investment choices (e.g. buying millions in Yen because they are momentarily miss priced vs. the Euro) to the point that I read that banks were putting the servers physically close to Wall Street, because the speed of light is a factor in beating out other banks. Also, I invest only in index funds, because I don’t know anything about investing. So I have no idea how much further it will go.
Rather than the folks who are being paid to talk, and to write things to scare us so that we will watch ads or buy subscriptions, I would look to the people actually designing AI architecture. Also, maybe google ‘Shoggoth and AI’ for how they view the future of AI.
That being said, AI could easily figure out the cost of a 3% increase over last year’s wages, assuming all other factors remained the same. And it could easily factor in payroll taxes and other factors which are generally too much work and have too little impact for me to bother including. But here are some of the factors that will be more challenging for AI:
Hours multiplied by wages - will hours go up, down or stay the same over a three or five year CBA? Even if AI is better at predicting the future than I am, I would trust a conversation with the GM over anything an AI could scrounge up. Is it likely that the State, City or local jurisdiction (i.e. airport, or hospitality sector specific are two I deal with now) will pass a higher minimum wage, which could cause salary compression. Will an AI understand how salary compression can cause you to lose valuable supervisors and managers? Can AI quickly do an accurate salary survey for your area and industry to see if you are already underwater in terms of wages vs. your competition? Can an AI explain (or would even think to do so) the difference between a 3% increase when your wages are already at minimum wage or will be, to help corporate understand that 1% of your increase wage going to happen regardless because of new wage legislation so this could be considered closer to a 2% increase? One challenge many employers have had over the past few years is costing out CBAs based on COVID-times hours vs. post-COVID hours. Some of my spreadsheets use 2019 hours because they are the closest to what we can expect going forward. How would you teach an AI to not use last year’s hours, and to use some other year’s hours instead?
Benefits - I will assume that AI can better predict what the benefit fund increases we will see over the next 5 years than I could. But could AI explain that this proposal has the lowest overall cost, but includes a defined benefit pension plan which could result in millions in withdrawal liability in the future? How would it value that? And how would that change where your company was providing a service, and you could be outbid and lose the contract (and trigger w/d liability) vs. a contract that you control and cannot lose?
Could AI understand the difference - going back to the example above where you are going to ‘any source’ hiring - between a union like the IBEW, who are dispatching trained journeymen, and another who had no formal training program? How would that affect the cost of the CBA?
In other words, there are a lot of subjective factors that go into some, but definitely not all, CBAs.
Last - if the AI told you the CBA was going to cost $1M, but you had no way to ‘look under the hood’ and see what its inputs, assumptions or formulas used - would you trust it? Despite over 20 years of experience, I have yet to cost out a CBA where a more senior exec (or more likely, a finance person at their direction) did not insist on looking through my spreadsheet to see how I came up with my increases over time. Which is as it should be - I’m not a finance person. But you can’t check an AI. It will tell you it’s correct, but it will lie to you - just ask that lawyer who directed an AI to write a brief who learned that everything was made up from whole cloth - even though it swore the cases were real. AIs lie too, just like people.
All of your points are excellent. Broken down, I think there are 'yes' answers to some, and 'no' to others.
I think (as I understand the technology [which is minimal, BTW]) is that it will be wholly dependent on the data input...the more data, the more 'knowledgable' AI will be and more likely accurate as well (in reference to the lawyer, for example).
As an aside, I am having the same economist who was on in May back on the podcast next week. He, by the way, stated several times that he hopes he is wrong.
Also, as a separate aside, I went into studying AI as a potential solution to labor shortage issues and population pyramids. It was the conversation with him that led me to go deeper and conclude this could be bad. (I am also trying to book a professor from UCLA on as well.)
If you're open to it, we should have a conversation offline. I am totally open to the idea that AI is a nothingburger.
Sure!
the idea that AI could interpret contract provisions or aid in disciplinary hearings strikes me as contrary to the idea that AI frees up people to work on "more complex" matters. Judges and arbitrators routinely disagree on the meaning of contract provisions, so how would AI solve this issue. Further, discipline is fact specific, involves witnesses, etc. AI could not possibly handle this.
Ryan,
Thanks for your comment(s).
To your point in the reply to Tor, as well as this one. AI (more than ChatGPT) "seems to simply compile a bunch of internet searches and put them in a cohesive summary" is fairly accurate.
[When I first started learning about AI, my analogy was 'Google on steroids.' However, I was informed that was incomplete. It is more and, in the future, will be even more than that.]
To that end, imagine a program that can compile all of the arbitration decisions published by the AAA over the last 100 years on a specific topic, as well as the Seven Principles of Just Cause. Then, imagine a program where a termination is input, along with the grievant's argument--even to the point of witness statements. As a former union rep (who's handled hundreds of grievances), if AI is able to grow to where "they" say it's going, I don't know that it is too far fetched to think it can be a useful tool.
I may be wrong...and, I do hope that I am. However, time will tell.