As AI takes over calculations, monitoring and troubleshooting, teams risk losing the hands-on knowledge that keeps physical infrastructure running. Keith Schafer's answer is that AI is useful but not truly intelligent, and over-trusting it erodes the experience you need when something surprising happens. Schafer, who has worked in natural gas and compressor reliability, spoke on The Artificial Intelligence Podcast. His view is that "As cool as it is, it's not truly artificial intelligence." His warning is not abstract: "It literally told me to drive off the side of a bridge." The same instinct applies when you let software replace the person who knows the equipment.
Why does losing hands-on experience matter in infrastructure?
Because the physical world produces surprises that AI cannot adapt to. In Schafer's work, people used analog oscilloscopes and walked around engines recording data on log sheets every hour. Now that data is automated, pulled into computers and SCADA screens, and visible from a cell phone anywhere. It is convenient and easy, but Schafer's point is that you lose a lot of on-hand experience when you do that. New people coming in have never seen an analog oscilloscope. Digital equipment has advantages, but you give up some things to make it easier and faster. The natural gas business has to move gas from point A to point B. If a company is not efficient, another gas company will move that gas cheaper. Safety and reliability come first. Schafer's rule is that "when it's 20 below zero outside, these engines have to run or a lot of people get cold."
What happens when you replace human monitoring with AI?
You remove the person who can adapt when something unexpected happens. Schafer's concern is that AI does not know how to adapt to surprises. The host's point is that projects want to replace a centralized person watching wells with AI, which makes him nervous when safety is involved. Schafer is hesitant for the same reason. The episode also ties this to data centers. Natural gas is being instrumental in the data centers because they are putting data centers where natural gas is. A previous company approved $2.5 billion for a new pipeline just to supply data centers. The transmission pipeline system and electrical grid need upgrades. Building that infrastructure means laying pipelines and putting in compressor stations. Regulations and right of way are drawbacks. The Keystone pipeline was canceled. Nuclear power is shifting, and small nuclear facilities have potential for AI centers. Natural gas replaced coal plants and dropped CO. But the knowledge to run all this is the knowledge at risk.
Are AI data centers making the knowledge gap worse?
Yes, because the infrastructure powering AI depends on expertise that is quietly disappearing. Schafer thinks the hype with AI data centers is probably overkill. Data centers need a lot of power, much of it for cooling fans. If someone comes up with other cooling, or if models get smaller, the power supply may be lower than expected. Data centers sign 20 and 30 year contracts. If they do not need the power, some may go bankrupt trying to cover contracts. His fear is that a data center coming right off the grid uses so much power that people cannot cool their homes in summer or heat them in winter, causing blackouts and brownouts. The host's point is that models are getting smaller, AI can live on a phone, and efficiency gains like a 70% reduction in RAM for AI are coming. Chips are valuable for 18 months. The hype cycle resembles virtual reality goggles. The strategic risk is building for assumptions that change.
What to do before you let AI replace a human operator
Put safety and reliability first. Schafer's rule is safety being number one, and the engines have to run when it is 20 below zero outside or a lot of people get cold.
Keep the human who can adapt to surprises. AI does not know how to adapt to surprises, and the host's point is that replacing a person in a safety role is where the hesitation belongs.
Preserve hands-on troubleshooting. Analog scopes, log sheets and walking the equipment built on-hand experience. Digital tools are faster, but you give up some things.
Follow up on OEM maintenance after repairs. In Schafer's expert witness case, a compressor caught fire after a repair because nobody followed up on work the original equipment manufacturer recommended, including re-checking and torquing. The contractor left, the owner did nothing, and the handoff failed.
Capture multi-generational knowledge before it walks out the door. The show notes put it plainly: the expertise powering the coming wave of AI data centers is the very knowledge at risk of being lost.
The choice is not AI or no AI. It is whether you keep the ability to notice when the tool is wrong, especially when the consequence is a fire, a blackout or a cold night.
Original show notes
Is AI quietly erasing the hands-on knowledge that keeps our infrastructure running? Jonathan Green talks with Keith Schafer — part of multiple generations in critical power and engine infrastructure — about what happens when we over-trust AI, why "artificial intelligence" isn't truly intelligent, and how the expertise powering the coming wave of AI data centers is the very knowledge we're at risk of losing.
Key Takeaways: • Over-relying on AI erodes hard-won institutional knowledge and hands-on troubleshooting instincts • "AI" isn't truly intelligent — blindly trusting it (like following GPS off the road) is dangerous in the physical world • The infrastructure powering AI data centers depends on expertise that's quietly disappearing • Safety and reliability come first — some systems simply have to run (engines at 20 below) • Capture multi-generational knowledge before it walks out the door
Notable Quotes: "The first time I used GPS, it told me to drive off [the road]." — Keith Schafer "As cool as it is, it's not truly artificial intelligence." — Keith Schafer
Connect with Keith Schafer: LinkedIn: https://www.linkedin.com/in/keith-w-schafer-19833a22
Enjoyed this? Follow The Artificial Intelligence Podcast and share it with someone who keeps the lights on.
Auto-generated transcript, 4,440 words. Timestamps link to the moment in the episode.
Is AI causing us to lose critical knowledge? We're gonna find out today's amazing special guest, Keith Schafer. Now, Keith, I'm so excited to have you here because you're part of multi-generations of something that's really critical to infrastructure, especially in America. And I know your grandfather started in the gas industry over a hundred years ago and your father was in the industry and you've been doing it a long time now, your son is in the industry. If all this generations of natural gas,
What do you think has made your family stick with this? Because there's one thing I know, it's my kids want to do something, but it's anything other than what I do. So how do you kind of have that consistency in your family? That's amazing to me. Well, like I said, I grew up on the pipeline with my dad, me and my brother. We'd always be with dad when he could take us out on the right away where they's laying pipe across the hills.
And I always thought that was interesting and loved to see me out there and outside working and enjoying the environment and seeing that done. And knowing that the... The natural gas business has always been a steady business. If you're good at what you do, you're always going to have a job. My grandpa started when he got home from World War I and my dad after World War II. They spent their whole careers in the business. I was lucky enough to spend 50 years in the business
now with the company and through contract work now that I'm doing now. I retired four years ago and, and, ~ doing a little contract work training, consulting, ~ just on my spare time last four years. So still enjoy it. Now when you started out, all the calculations were done on paper with paper and pencil. And I think that we have started to move away from trusting our ability to do math in our heads and then it was calculators.
And now it's having an AI do the math for you. And we're kind of shifting away. And this is kind of how I feel like I know how to read a map. Whereas like people five years younger than me never even seen a map because they've grown up with GPS. And first time I used GPS, told me to drive off. It literally told me to drive off the side of a bridge. And I was like, I don't think I'm going to trust you.
So I have this paranoia about it, but we move to newer newer tools and we kind of forget that foundational thing. And how are you able to have trust in your calculations or what do you think is happening where people are shifting away from trusting their own ability to do math in their heads? Even now, at a lot of places, you can't even make your own change. It figures out the change for you. Yeah, I feel lucky to get started in this
business when we still had that. A lot of the new guys coming on now, they've never seen that. My 50 years has been in compressor reliability work and we use a lot of test equipment and things that I was teaching a class last week and we were just talking about before I used to use the analog oscilloscopes and The guys that I've trained in, never even heard of an analog oscilloscope. It's less used one. And now everything from the 90s,
late 90s up is all digital equipment. And there's some great advantages to it and there's some disadvantages to it. you give up some things to make it easier and faster and easier to use when you go to some of the new technologies. But after you do it without it for so long, You don't know that you missed it. And I can still talk about some things that we used to be able to do with analog that digital is just now catching up to.
And like you said, everything was, ~ we walked around our engines and equipment, recording data with a log sheet every hour on the hour for long time. And then now all that is automated and pulled into the computers and our SCADA screens and we can see it anywhere in the world that our pressures and temperatures and things that we want to collect on an engine or a piece of equipment anywhere in the world from your cell phone now. And it makes it convenient and easy,
but we lose a lot of on-hand experience when we do that sometimes. Yeah, I see my kids, I have five young kids and the way they deal with technology is so different to Like I'm used to if a tape or video won't play, you pull it out and blow on it. Or you slap the side of a TV and that's actually made things work before and we don't have kind of this tactile experience. So we're starting to see this knowledge gap where everyone from
my generation was like college is the solution and now the average plumber in America is over 60 years old. So there's this huge gap when they all retire. It's going to be the seller's market for plumbers, right? Because it's like, there's only one in town. You can charge whatever you want. And we're seeing a lot of these types of things happen with these jobs. And I think that a lot of people the past 20 years thought that we were going
to be all switched to windmills and solar panels by now. But now with this massive spike in need for energy from AI data centers, like every single day, there's an announcement of a new data center and they use like insane amounts of power, like more than an entire city and sometimes more than a state. Are you already starting to see the need for energy spiking in the projects you're working on and in all of the different natural gas needs and traditional oil needs that
are coming because we're starting to see, we're building data centers everywhere and especially want to build a remote locations closer to the power center. So with this gap, is there going to be an even bigger need for your industry going forward. yeah, the natural gas business is being instrumental in the data centers because they're starting to put the data centers where we have the natural gas now. A previous company, TC Energy, I just saw last week that they already have
approved $2.5 billion for a new pipeline just to supply data centers. The transmission pipeline system and the transmission electrical grid across the US has got to be upgraded because we don't have all the energy where they want it. And that's how you, it takes a lot of infrastructure to get it there. We can put in a, where we can get the gas in the Shell regions in Ohio, Pennsylvania, West Virginia, Texas, but we still got to get that.
to the sources and that's laying a lot of pipelines and putting in lot of compressor stations that ~ is a big demand for that. That's something I really wanted to dive into, which is that a lot of people think you push a button, the power turns on. But what does the AI industry really not understand about what it actually takes to get the gas from a well-headed Pennsylvania to a data center in Virginia? How long does it take to build the pipeline?
How much maintenance does it need? And how complex is that process? What's made it so complex a lot is the new regulations of laws with the environmental issues that we have all the time. So that's one of the biggest drawbacks of, and just getting right a ways to build these power lines or natural gas stations or pipelines. So a lot of it is environmental regulations that, we've to jump through. If you remember the Keystone pipeline just a years ago,
we had that and it would have been pumping a lot of oil that we need right now if it wasn't canceled. So that was canceled just political and environmental issues that they thought would have. And if they would have built it, it would have really come in handy in our oil prices right now, our gasoline prices. Yeah. So one of the things I've noticed is that for a long time, I think we haven't built a nuclear power plant in America since the late
70s or early 80s. just didn't make it, ban it, but there's been a moratorium because of the Three Mile Island incident. Right. suddenly now everyone seems to be okay with nuclear because like, if we're going to take enough power for nuclear power plants, need nuclear. oh, if it me AI, then I'm okay with nuclear. So like I've never seen such a fast shift and how people felt about something with environmental. Are you noticing the same kind of shift about natural gas or
oil that suddenly people are like, well, we really need, especially right now with prices skyrocketing for other fuel sources, are people starting to become more relaxed or like shifting their environmental views? The people I deal with has always known natural gas and oil is the way to go. Nuclear power, I've always been a fan of that, if it's done right. And now the micro stations, the small nuclear facilities, I think that's got a lot of potential for leading into the AI centers that's coming up.
yeah, there's what going on with the natural gas. I'd like to say, the environmental side of it has always been against the natural gas and oil business. But when they started looking at the biggest reduction we've had in the US on CO is because of natural gas taking the place of the coal burning power stations that they've displaced. So natural gas is replaced a lot of coal plants over the last 15 to 20 years. And that's dropped the CO in the US by numbers they never
even imagined a few years ago. And we're still producing more because we're using natural gas instead. natural gas is a clean burning fuel, but it just gets a bad name because they say it's a fossil fuel and it gets in that category of oil and coal. It seems like there's a lot of misunderstanding about how natural gas works because either you have a natural gas stove and so you're used to having a pipe to your house, which I had when I lived in Tennessee,
or you imagine that it gets burned right where you drill it and then turn electricity and sent through the wires. So people are, I think there's a pretty lack of understanding. And also everyone tends to be like 10 or 20 years behind. Like everyone's really just now talking about fracking, which I know has been around for a long time. You kind of mentioned, and I've heard that like, it's now about horizontal drilling, like you turn down and then take a right
turn to get to the natural gas or to what's down there. How is the industry changing? Why do think there is this kind of delay between what we're doing now and what people hear about, which is from 20 or 30 years ago? Yeah, 20 years ago when they discovered the shell, when they really discovered that the shell plays in Ohio, Pennsylvania, West Virginia have been there for millions of years. But the trouble is drilling a mile to two miles deep and putting
a straight hole in the ground wasn't very economical. the invention of that horizontal drilling, now they can drill a mile deep and go out in that right-hand direction for a mile. Then take another hole straight down in that same hole and go another direction so they can fan out a lot of different directions. And now they have access to a lot more natural gas that would have taken thousands of holes in the ground. But now they can do it just one hole.
That was the technology that really got the boom going. The fracking is the term that everybody that wanted to be against our business said was destroying the environment and having poisoned chemicals down the wells. And that was what they drove on. But the whole technology was the horizontal drilling that really got the shell play moving. fracking is something they've been doing 100 years in the oil and gas business. And you never even heard of it until it started in a natural
gas business or with the Shell plays. So it's a lot of hype on all of that, but the drilling, now the technology with that, we can get a lot more gas out of the ground. And then we build the infrastructure. when I was heavy, working with TC Energy, we was building pipelines and moving it from West Virginia. We brought gas out of Texas and the Gulf of Mexico for 65 years and brought it up to West Virginia, Pennsylvania, for gas distribution.
Now with the Shell plays, we are now sending 3 billion cubic feet of gas a day. down that pipeline going to the LNG facilities in Louisiana and Texas. That's just one pipeline. The whole shift of where we're getting gas from and where we're sending it to has changed dramatically over the years. How has technology changed over your kind of journey? I've been approached by some projects where they want to. And I know we have skater, which is amazing, where one person is centralized place can
watch all the different data. So from a bunch of centers, right? So they're watching all their wells from one location. But now they're talking about replacing that person with an AI, which always makes me nervous when I get approached by projects that are safety because AI is. As cool as it is, it's not truly artificial intelligence and it can never, it doesn't know how to adapt to surprises. And if there's one thing I've learned, it's that no matter what happens, it's all, there's a surprise somewhere.
And like, that's kind of one of the areas that worries about me and we're constantly thinking, let's replace everyone with AI, let's replace everyone with AI. How are you kind of seeing these shifts in your industry and where are the good places where it's okay? And where are the places where you're a little bit hesitant like I am? I'm little bit hesitant because I'm thinking the hype with the AI data centers is going to probably be overkill because we are building massive data centers and
say we need all this power right here. And we are trying to supply for them. And I think after a few years, the power supply is going to be a lot lower than what they think. And we're going to have all this infrastructure. moving electricity and gas to these data centers that's not going to be required anymore. So I think there'll be a lot, I think we're over killing a lot of it right now that with the technology coming
you know, what they say we need all this power for the data centers. Do you know why they need all the power? A lot of it is just for cooling fans to cool from the heat that it generates. So when you, if somebody comes up with technology to get some other cooling besides like what they're doing, putting a lot of these data centers in water and under the sea and stuff to not have to supply all their cooling capacity for that.
So there's things that I think right now is going to be way overbuilt on the power side and that's going to cost some money for Hmm. a lot of people coming up. And then my fear is building a data center and then coming right off of the grid and using so much on the grid that people won't be able to cool their in the winter time or summertime then. We'll have blackouts and brownouts because the data center's taking all the power.
Yeah, I think that's when people will turn against AI when it's air conditioning or AI suddenly AI is not as cool anymore. But very quickly. The thing about these data centers I think about a lot is they make two big assumptions, which is that everyone needs more more AI and that AI will not become more efficient. like the models are not getting bigger, they're getting smaller. Like what everyone really wants is an AI in their that they don't have to be online,
that just works. And that means it needs no data center. And I already have seen there was a big like technology recently came up from Google that massively decreased almost by 70 % how much RAM and AI needs. And we're going to see more and more of that. We always see that, right? We have a big technology, then we find an efficiency, then we have a new technology, we find an efficiency. And all these data centers, the problem is the chips already are only
valuable for 18 months. And then it's like, you're using last year's chips gross. It's such a fast turnover that I think you're exactly out into something that We're assuming that the chips in two years will use the same cooling, will need the same power, will need the same size. And it's why we're constantly building new and new data centers. And it's just, it reminds me of the hype cycle a few years ago about, you know, everyone said they were going to wear virtual reality goggles
and they built all this stuff and turns out no one wants it, Exactly. right? The technology shifted like, there's no way I, I have five kids. There's no way I'm putting on goggles and turning my back on them. Like that's crazy. That's why, let alone we all, we're all wearing goggles to watch a movie. So we're all together, but all alone. And so sometimes I do think that we get excited by this idea and data centers, it's a great business to
be in right now because everyone will give you tons and tons of money, but they all, know, 18 months later, it's like everything we spent millions of dollars on is worthless. What do we do? And it's a constant cycle. So I think that's something that's really important is to look at. more strategic and more long term. there are some, what we're going to see is like this big spike in energy demand and then a big dip, right? We're going to have all this left or energy.
actually we don't need these data centers because now it can live in your phone. Like two years ago, you could not put any AI in your phone. And now you put a pretty good one in there and it's going to get better and better to where, we don't need these things. And so these assumptions we've made are kind of going in the wrong direction. But I'm also interested in, yeah. So to, well, I was going to say the very thing right
there to build a major pipeline or transmission line to a data center and supply gas, they have to guarantee that we're going to get paid. So they'll sign 20 and 30 year contracts to take this much power from us. And when they do come that and they don't need it, some of these are going to go bankrupt trying to just cover their contracts. So there's gonna be some things stirred up that will not be able to sustain themselves. Yeah, I think that's exactly what happens,
right? You get too excited and you're so busy thinking about the upside, they don't think about the downside. ~ One thing that really interests me, I know you had this amazing experience as an expert witness in a case in New Expo. I'd love to hear about that because lawsuits drag on for years and years, but you were able to settle that thing pretty quickly. Can you talk a little bit about that story? no, I'm sorry. I didn't hear the first part.
about your, you have an experience being an expert witness. I would love to hear that story. ~ it's just with my experience around equipment and engines, compressors, had a group of lawyers contact me about wanting to be an expert witness on a ~ fire that they had. And I thought, well, they got the wrong number. And he said, no, he wanted to talk to me about it. And I said, what? found out a little bit about what they had, had a compressor out Midwest that caught
fire and burned up in the middle of the night. It caused a lawsuit between the pipeline company, the packager that supplied the equipment, and then the contractor that worked on the equipment. they had the fire happen, so... Each one was blaming each other for the fire. And when I went to the site and looked at a unit that had been burned, completely burned up ~ for two years out in the field, I could tell pretty quickly ~ just about what happened and how it started.
And once I told the lawyers what my opinion was, they... ~ They took it to the other group of lawyers and said, you don't have a case now that this is what happened because of lack of the ~ companies kind of dealt with each other with a handshake because I said, what's the scope of work that you had? And they said, we didn't have a scope of work. We just agreed to fix it. And I said, okay. they, they, nobody knew what the
other group was doing. And when they said, Okay, we fixed your engine, we're done. They paid them and they left. And 30 days later it caught fire. So I could tell that the handoff was probably the biggest problem because they didn't follow up on work that's supposed to have been done by the OEM, original equipment manufacturer. They recommend after you do these repairs that Certain things have to be re-checked and torqued and tightened up. And nobody did that because the contractor said that when
we finished, we left, we didn't pass to do any follow-up work. And the company that owned it, they didn't do anything of their self either. So ~ the lack of following up with the equipment manufacturer to see what maintenance need to be done, they didn't do, and that's what failed. It caused a fire. So I, after I went and investigated it, talked to the lawyers that hired me, told them everything. was about two months later, they called me and said, they settled down to court.
We don't have to deal with all that now. So it didn't have to work. So it's fun and interesting. I don't know if I'd want to do all that work all the time, but it is another. interesting job that I had in last year that was a pretty good, unique situation for Yeah, that's really exciting. I've always wanted to be an expert witness, Yeah, yeah. Yeah, I didn't get to go back to the trial, but with what we did, they can settle out of court then.
Just one last question as we wind down. This has been so exciting. Just curious when a gas company decides to adopt a new piece of technology, whether it's predictive analytics, AM monitoring, and or a new tool, what does that adoption process look like inside the industry? Because from the outside, people perceive energy is very slow to change as a slow moving industry, kind of like healthcare or education. Well, we, the industry, we work with different companies across the whole footprint, all over the world,
really. ~ And we see what our reliability is. So we know we have to deliver gas from point A to point B, and we have compressor stations that move that gas all across the country. And if we're not efficient, company are... another gas company will get that price to move that gas cheaper. So we have to keep our equipment as efficient and maintenance free as we can. the technology that we develop, the other gas companies develop, and we have contractors that ~ work with us that
we deal with as far as equipment manufacturers that build this equipment on what we tell them we want. ~ then we have to hire people or contract people out to do the test work that we want to be, ~ say that we're safe and reliable to do this. And a lot of last thing we want is other failures, other pipelines and anybody getting hurt in our industry. It is a very dangerous business, but if you do things right, you can move
it safely all over the world. ~ It takes everybody doing it right. And ~ you don't want to give up ~ something that's quick and easy for we spend a lot of time making sure that our equipment is running properly, not only to be efficient so we can move it as cheap as we can, but we got to do it safely as we can. So we got to keep those things in focus there, safety being number one, and then everything else works down on how we
how we move our technology and distribute across each of our companies to ~ get all of our equipment running when we want it to. And we have to have it reliable because when it's 20 below zero outside, these engines have to run or a lot of people get cold. Yeah, that's amazing. Thank you so much for sharing all this information. I know that our audience is really going to love this. So thank you so much for being here today, Keith, for an amazing episode of the
Artificial Intelligence Podcast. Good job, son.
More on ai in manufacturing, energy and infrastructure
