AI can tell an operator what a plant will likely produce 24 or 48 hours from now, but it cannot be handed the decision on a high-production, high-security process. John Hooker, a control engineer at Statistics & Control, told The Artificial Intelligence Podcast that the workable split is to let AI handle the questions that are not critical and to check its answers against a physical model before it touches anything that is.
Where does AI actually help in an industrial plant?
It helps with options, trade-offs and forecasting, not with the final call. Hooker describes AI as not quite at the decision-making level yet, but useful for showing operators "here's the options, here's the trade offs." His company builds physical simulations from real physical models and real-time data, and consults them to see how a change would affect operation. AI is fast at generating that information, and that speed is where it earns its place.
Why do general AI tools fail at precision work?
Because they were built for creativity, and industrial processes need the opposite. Hooker points out that AI is a broad term, and that many people take a large language model built for one type of use case and port it into manufacturing or scientific work, where creativity is the last thing you want. Front ends and wrappers do not change that: the underlying technology still was not purpose built. What follows is risk.
How do you keep AI from being wrong about a process?
You compare it against a model you already trust. Hooker's team spent years building trust in its simulation, adjusting coefficients and checking that the model was set up correctly, before adding AI alongside it. When a question is not high security or high production, AI can answer it directly. When it is, the answer gets tested against the physical model to see whether the two agree. That comparison is the safety check.
What breaks when you assume every site is the same?
A solution that works at one station can make another one burn more fuel. Hooker describes a steam generation project with four stations. He surveyed one, was told all four were the same, and built a system that optimised the first by adjusting water flow. Only the first station had variable speed water pumps. At the other three, operators could not make the adjustment the system asked for, so they burned more to compensate. The lesson is that you have to enter correct information, and that nobody, including the people on site, knows all of it. Diagrams show valves that are not there. One department installs something and does not tell the people responsible for digital solutions. A valve is turned by hand and never shows up in the data.
How do you know the data you are feeding it is right?
You check it, because AI will trust a sensor that is obviously broken. Hooker gives the example of a pressure sensor reading a million when the process normally runs between two hundred and four hundred. A person recognises that as a calibration or engineering-units problem. AI takes it as the system's word. The same applies to a control system that is not bringing data in correctly: once you start making assumptions to fill the gaps, the model is only as good as those assumptions.
What to do before you hand a process to AI
Know your workflow and your process first. Hooker's answer to what a company needs locked down before implementing AI is that it needs to know the process behind what it is doing and how it is doing it.
Sort the work by stakes. Let AI answer questions that are not high security or high production, and route the rest through verification against a physical model.
Ask how the AI reached its answer. Hooker warns against trusting output blindly. Ask how it got there, and be aware that a model will often find one solution and stop rather than lay out the trade-offs.
Fix the data before you scale. Reconcile diagrams against what is actually installed, check that the control system is bringing data in correctly, and watch for sensors that are badly calibrated or in the wrong units.
Keep someone who knows the process in the loop. "You need somebody who's very familiar with that process," Hooker says, because the machine cannot yet explain why a reading is odd.
Test the model on the hard cases. Hooker asks new AI models to write a proportional integral derivative controller, and finds they forget the endpoints and the question of how small an error needs to be before it is worth correcting. In oil and gas, decimal points are not critical; in pharmaceutical and food work they are.
What does "half robot, half person" look like?
It already exists in the form of remote startup support. Hooker describes commissioning work during COVID where an operator on site wore a camera on his head and live streamed to him while he sat in Des Moines and the plant was in Germany. He told the operator how far to open a valve, when to pause, when to move on. The session was recorded, so the same startup can be reviewed before it is done again, and the lessons stay with the company. Refineries that once needed a thousand people around the clock now need about two hundred and fifty, he says, with monitoring moving to a central hub. AI's version of this is standards and checklists built into the work, checking that every tag is named correctly before a turnaround. The open question Hooker raises is whether a site will accept being told not to run something yet because it does not meet the standard, when what the site wants is to start up safely and soon.
Original show notes
We think of AI as digital — but can it run a real, physical oil field? Jonathan Green talks with John Hooker of Statistics & Control about AI in high-stakes industrial operations: where automation genuinely helps, where you still need a human who knows the process, and what "half robot, half person" operations actually look like.
Key Takeaways: • In high-security, high-production environments, AI augments — it doesn't replace the expert • You still need someone who deeply knows the process to supervise the automation • The future is "half robot, half person" — humans and AI running operations together • Real-world signals (fuel flow, water, pressures) are where AI earns its keep • "Adopt AI" means knowing exactly what to hand it and what to keep human
Notable Quotes: "It's almost like it's half robot, half person." — John Hooker "You need somebody who's very familiar with that process." — John Hooker
Connect with John Hooker: LinkedIn: https://www.linkedin.com/in/johndhooker Company: Statistics & Control Inc (West Des Moines, IA) Website: https://stctrl.com/
Enjoyed this? Follow The Artificial Intelligence Podcast and share it with someone in industrial operations.
Auto-generated transcript, 5,881 words. Timestamps link to the moment in the episode.
Can AI actually run an oil field? Today we're gonna find out with our amazing special guest, John Hooker. Now, John, I'm really excited to talk to you because we often think about AI as something that's like digital or lives in outer space. But when people say the cloud, they really think it's in the air. But there's actual physical infrastructure and actual machines where even the tiniest of stakes makes a really big difference. Like when you're drawing, I mean you're
like making an AI picture of your friends, an extra finger is no big deal. But being off by even like one percentage point or one millimeter can cause precision instrumenting to crash. And so there's no room for error in your industry. How can AI enter an industry when AI's kind of the double edged sword is that you never know what it's gonna do? It's unpredictability is both its benefit and its greatest negative because you need that core precision. Yeah, absolutely. That's a great question.
~ one of the more amazing things about AI is just, you know, right now it's in the stage of where people are using it more and it's getting implemented into there. It's not quite at the decision making level yet, but it's helping a lot of different operators with decision making or saying, Hey, here's the options, here's the trade offs from them. And one of the things that people are doing quite a lot of and that we've seen in the industry quite a lot of,
although we're a little bit biased since one of the things we do is we build our own physical simulation based off of real physical models and real time data. One of the things that we're able to do is we're actually able to consult our simulation and say, hey, if we make these changes, how will it affect operation? How will it do that? And one of the things that's exciting about AI is just they're very quick to get answers to. Get to generate that information that you
need of, hey, what's my production going to be running at 24 hours from now? What's my production gonna be running at 48 hours from now, or even ~ a month from now, we'll probably be a little bit better as data improves and ~ AI gets better. But one of the things that simulation does is especially when you start creating a physical simulation, it's quite a lot of computational power, it's quite a lot of resources that are involved in making sure you set it up correctly,
you have all the coefficients adjusted right and that we trust the model. So building the trust with that model, we've done for several years. And now it we're kind of adding AI into that as well and saying, Hey, okay Let's get answers from AI if it's something that isn't quite as important or something that isn't, you know, a very high security issue or very high production issue, then we can go ahead and answer that. But other situations it's like,
okay, well, let's ~ try and test it out with this physical model and see if we get the same thing. And so it's kind of being able to add that verification back on that in there. So like I said, right now it's kind of driving towards okay. we're wanting we're asking it what can we do in a certain type of situation or what are the outcomes based on what's happening now, and then how well is that answering that?
And so by being able to compare that it helps put a little bit of a safety check in there. One of the challenges is that AI is such a broad term. And so we see a lot of people trying to use like ChatGPT or Cloud or just a large language model, which is built for one type of use case and ported over to manufacturing or scientific use cases where creativity is the last thing you want. You want precision. And even when you see ~
tools that are front ends or wrappers, they're still the underlying technology still comes from something that's not purpose built. And so you're le entering or like leaving this room for risk or danger. How can a company kind of in operations, in manufacturing, in the physical world may ensure that the tools they're looking at are built for precision rather than creativity? Yeah, that's also a great question. One of the things that's happening right now is most people are building operational knowledge inside that large language model.
And so basically right now you have a lot of APIs that can connect to a lot of different data. ~ also companies are very interested in the security of their data as well. So a lot of them are building their large language models ~ kind of in house or being able to ~ use that and then they have APIs that will connect to data, and then one of the things that they try and do is right now we're
more in the okay I have a sensor that went down. ~ why did it go down? So then the large language model can search through a database or have former of from individuals that work in that process that have put it in there and they can get examples of all right, well maybe it could be ~ the wiring is default. So let's go check the physical wiring of it. It may be that, it's connected from one software to another or
it gets transposed this way. So let's check and see if it's ~ you know configured correctly going from the field to the main kind of business ~ level enterprise where that data sits now. Or it's one of those where it could be, hey, it hasn't been calibrated in six months. So maybe we need to go take a look at the calibration and see how if it's off or not. And so that's one of the things that people are being able
to search and come up with. Okay, I need this checklist that normally you would need either an experienced person to have that knowledge of. Or you would need somebody who's very familiar with that process. Well now you can just go ask the machine about it and then ~ be able to have an answer for that as well. In terms of manufacturers trying to make it get a little bit better or make it more interesting, one of the things that especially with sensors
and with ~ monitoring processes or monitoring ~ type applications, they're you know, taking a look at a huge statistical range and basically being able to save all that data inside a database to where they can do examples of it. And one of the things that I expect from manufacturers in the future is they're probably going to develop ~ either a certain API to wrap into the customer's AP or the customer's AI either agent or ~ their large language
in order to help out with ~ text in the field or help out with people that are doing a lot of the decision making or monitoring a lot of it to where they'll probably need less and less people. Actually one of the things that I've seen a lot ~ in refineries is refineries used to be these big huge operations. And when I first started in the business in about two thousand ten, they were, you know, normally h had five hundred, six hundred, a thousand,
twenty five hundred people man them constantly, twenty four seven. Well now automation has increased so much and even ~ people who don't want to go sit in the middle of Dickinson, North Dakota or don't wanna go to Salt Lake City, they wanna have their intelligence and kind of their data solutions in a central hub at headquarters, they can monitor all those facilities. there and so there's less and less and less people on site and there's not needed anymore. So operations that
needed a thousand people about fifteen ago only need two hundred and fifty people. And so I assume that it's also going to be able to go what down or just increase the efficiency essentially with ~ people who work there ~ in that. And so that's how I'm curious to see how manufacturers will do it, where it will just increase the efficiency of what's in there and you don't need people to look at data essentially you can have your AI look
at data and then it can send it up ~ to either ~ central or it can send it to the main plants person you know SME that's there for the process and then they know what to do with it essentially or be able to use that information or data ~ a little bit more effectively or just a little bit faster to prevent either shutdowns or prevent ~ emergency shutdowns, things like that. So that's probably where I see manufacturers move a lot towards it.
In lot of science fiction books and na that I've read, they'll have engineers wear goggles that can see what they're seeing and then it like tells them what to fix. And it's kind of a way to idiot proof repairs. And I've talked to a lot of companies, that's kind of their dream that the AI kind of guides it and then the person's hands are falling, almost like it's half robot, half person. The person becomes the hands of the AI.
How far away do you think we are from then? Do you think that's a realistic future? Because I feel like the downside is that you the person is doing stuff but they don't know what they're doing. That's that it's a very interesting ~ case in that 'cause one of the things that happened with us, so we do a lot of ~ controls as well and we upgrade a lot of controllers, we upgrade a lot of plant processes, different things like that. And one of the
things that we were doing a lot of with COVID was we had to do stuff that was normally meant to be on site. We had to figure out a way to do that remotely. And we've even implemented kind of a ~ the ability to do site acceptance tests where the operator was wearing a camera on his head. And he was just live streaming to me and me alone. And we were able to see what was going on and help them start up,
whether it be a turbine and a compressor, or ~ in this case, I think it was a steam turbine that we were starting up. But ~ it's one of those that was really funny because here I am ~ sitting in ~ the middle of Des Moines, Iowa, and it's at a plant in Germany, and we're saying, Okay, you know, let's open up the valve this much and all right, just pause for a second and ~ let it go through, let it cycle.
through and then we'll move on to the next step. So it's things like that that I think ~ will help out. And another nice and important case like that is they actually had a video recording now of us starting up something. So say for example when They are upgrading their system again and hopefully they choose us in twenty years to upgrade the system. I can ask them, hey, I know we actually recorded something ~ on that. Again, do you still have the video file ~
of that? And it'll be nice just to kind of see ~ how things are started up or what they did there. And it's also just a nice case that you can review that before you do it again and then also kind of continue to take with the lessons learned and lessons ~ available from that. And I'm very curious to see. So again, think of that was only just one piece of albeit very expensive equipment that we were doing. But if AI can just be sitting there
monitoring everything and watching it and you're like, Hey, when we connected a sensor into whatever h a honeywell ~ control system you know, how did we do that and what's the best way to do that? And then also another nice thing that they can do is hopefully companies have standards and different things like that. It's not always ~ it's hard to audit every single standard and every single thing that people have, but it's also one of those that you can build
in those standards inside the audit and build in those checklists or AI can build in those checklists to help either make the work go a little bit faster or just make sure that you're using those standards in order to make sure everything works correctly. So well, it'll be interesting to see just how one of the things that I'm very curious about is okay, say for example, a company implemented that particular solution where they have standards and they say, Hey, go out to the site,
here's our control systems. Do we have every single tag named correctly? And how much it will flag? And also, especially when you start doing turnarounds and startups, one of the things that matters quite a bit is time, especially, and so I'm curious with how much effort and energy people will go into either fighting that standard or fighting that thing because at the end of the day we want stuff to turn on. But it'll it'll be interesting to see like,
this isn't up to standard, so don't run it yet, or hey, you better find a way to change it. So that way we can make sure that it stays within our standard. So it'll be very interesting to see how that particular conversation goes because most of the time as a consultant the site is just like, hey, we just wanna start it up and we wanna start it up safely. We don't really care whether okay, the tag isn't named the same, or we don't
really care that you know, something is started up just a little bit differently from what we have at a different site. And that may be that they just live with that because of it, or they may just say, No, we need to make sure that it actually gets standardized so that way we know how to do it and ~ we it's repeatable in the future for future sites that we do it at. So It'll be really interesting to see what goes from there, from that.
So one of the things I've noticed in the work I do, which is on the softer side of industry, not the hard side you're in, the mechanical side, is that a lot of times there's this idea that AI can figure out the process and then accelerate it. Whereas for me, whenever a client has like a really detailed SOP and they're like, sorry, it's so detailed, I'm like, no, the more detailed the better. I don't care if it's a thousand steps.
If you have an instruction manual or Bible that I can follow, it's so much easier than the guesswork. And I find that it often leads to these unreasonable expectations or this belief that something's possible. It's not possible. It's like I can do a process faster, but if the process it's just gonna crash faster if it's a broken process. I'll accelerate whatever you're doing, but if you haven't gone through that testing and iteration, figuring out things, just as you know, like compressors work different at
different elevations or if it's raining or You know, you have to have these different things. It's very different in North Dakota versus Texas, because one's hotter and one's colder and things like that do matter. And the computer won't think of those unless you mention it. So I find that that's a really big challenge. Are you seeing that in the manufacturing industrial side as well, where there's this belief, well, the AI can figure out the process and then design it and then accelerate
it all in one go? Yeah. ~ one of the things that we do a lot and probably our customers are not as big of a fan of us for it is one of the things that customers do is they hire us because we're the experts and that's great, but also they're the ones who work with the process a little bit more. And so again, a lot of times I kind of have different ways of saying, Okay, we can do it this way or we can do
it that way. Here's the benefit of one way, here's the benefit of another way, here's the con of one way, and here's the con of another way. And it's looking at those trade offs and the the thing with that is is that is literally just from experience. So it's I I I think I've probably commissioned more than 60 projects in my career. And it's just from being on site and doing those 60 projects that it tends to start to grow. I think our
company has commissioned more than 500 projects over the course of our company history. And being able to have that knowledge where I can go to other people and ask, hey, have you ever ran into this issue? Or hey, we're looking at two options here. Which one do you think is really great? I think one of the biggest ~ issues that people have is when they ask AI a they don't take into account a lot of either those trade-offs or AI is very
bad at asking that. They just find one solution and say, Hey, this worked correctly, let's just go ahead and do this. And so that's one of those where it doesn't work out ~ very well. A a great example that ~ I have of something similar to that was we were actually doing a steam generation project. And one of the things that happened was they said, Hey, implement a steam generation project here and we'll be great. And one of the things they had,
they had four steam generator stations essentially. So I went and looked at one of them. I did a site survey for one of them and they said, Just do one, all four of them are the same, don't worry about it. And I knew and unfortunately I was too inexperienced to be like, No, we should do all of them or we should at least You know, take a look and talk to all operators and do that. And it was one of those where we developed
a solution that was really great where it did optimize the steam generators. We had fuel flow, we had water. you know, we we put out the set points for the fuel flow to get a certain steam quality to generate that. Hopefully we'd balance the steam generators based on efficiency, it'd be great. And one of the solutions that we had was in the first steam generation station, it was their ~ they had variable speed water pumps, and so So but in all the other steam generation stations,
they didn't. And so that's one of those where we implemented a great solution, but it was not very effective. for the other three steam generator stations. So on the first one that we did, we can say, hey, you can save a lot of money. And then we just kind of, you know, transpose that. And of course the other people did the same thing. And then when we went to go compare, like, okay, this is how we're running versus this is where we expect it to be.
And normally one of the things that we try and do is we try and close our target towards what's actually happening. And once we close that gap, we can see like, all right, cool, you're as optimized as possible or you're optimize you're where you want to be essentially. And in this instance, it was actually kind of the opposite happening where they were burning more to adjust because ~ our system was looking at, hey, all you gotta do is adjust your water flow.
Well, they couldn't do that. So in this instance, I see ~ one of the things is the you know the you have to enter in the correct information or you need to know all of the information available, which is a struggle even with people who work in the field and work on these things because of either just the way that corporations are set up or the way that people are set up. ~ It's just we you don't necessarily know everything and sometimes a lot
of data gets omitted or we even find ~ when we ask people like, you know, I see on this P and ID diagram that you have this particular valve here. ~ however, when I did a site survey, we don't have that. And so it but this was on a date, you know after my site survey. So did you actually install a valve or did you remove a valve or something like that? So it's things like that that we run into a lot.
And especially when you start getting involved with larger companies and different departments, you know, maybe process wanted to install something. So they initiated that, but they didn't talk to say for example, ~ the person who's responsible for ~ digital ~ solutions in their particular company. And so the people, the digital people that we're talking to don't know anything about that. And so it makes for something to where is just you that's why one of the reasons why I like our simulation is
in our simulation I can see that like, hey, we have something different, you know, why is that? And then we can say, we missed a valve there, or it's another one of those situations of ~ a valve ~ turned on and it just didn't tell because it's not digitized, somebody manually went out. did that. So it's ~ being able to bring everything up into kind of the 21st which is difficult because it's expensive, and also ~ being able to make sure that
everything is kind of standard with the data that you're bringing in ~ as well. So I think that that'll be an interesting thing to see in the future of how that progresses and grows. Because when we first started out, one of our big problems was with our simulation, we needed data in the field and we were pretty good about if we we kind of knew what data to expect. But it's one of those where people are like, yeah, we have some missing parts or they
think that they have their data in there but they screwed up the SCADA system and the SCADA isn't bringing in data correctly. Well now we start have to making assumptions for that. Well, how good is your model? How good is your simulation when you start making assumptions? But there's not much we can do about it because they said that they had data for it, but now there's not actually any data or it's coming in as bad data.
So it's being able to kind of ~ help out with that. And that's another thing that AI, I'm not you kinda need to train your AI on like, hey, we have ~ you know, say for example a pressure sensor and normally that pressure sensor, the process is only between two hundred and four hundred Psi. Well that sensor is reading a million. So I'm a guess that that's either badly calibrated or there's some kind of engineering units wrong with it. But does AI know that? Well no,
they're just like, well we'll just that came in from the system and we said we can trust the system. So that's ~ where it is. So it's it's an interesting aspect. For the future. So this is the biggest push I've ever seen for everyone to adopt a technology. So I remember when everyone's like, you need an email address. And then I remember when they were telling every restaurant needs a website, and they're like, Why? And then you I also remember when everyone needed a MySpace page,
which you don't need anymore. Like you couldn't be a band without a MySpace page for about seven years. And then we've had these pushes. This is the first time where suddenly everyone's trying to adopt this So every industrial company and every web every restaurant, everything in between is like we need to adopt AI. And they're not really sure what it means. And part of it is ~ there's this FOMO or this everyone else is doing it. We don't want to get left behind.
What from your experience, like what does a company need to have locked down in place before they start implementing AI? Like what do they need to have absolutely right? Because otherwise AI is just going to make it worse, not better. Yeah. I I think that they need to know their workflow and their process ~ you kind of nailed down. And one of the things that we see right now, as I said, is everybody's trying to ~ you know, just feed either their large language
model more data 'cause that's one of the things is it always needs more data and it always needs more clarification, more detailed type ~ aspects of it. And I I think one of the things that people need to nailed down especially is the how it comes up with answers for their solutions. So again, I'm a control engineer and one of the things I love to do with the latest and greatest AI models is I ask it to make a P I D for me.
Just program a simple C sharp ~ P I D program. So again, it's a proportional integral derivative controller and basically it's all throughout process industry and one of the things that it does is it normally comes up with the first case of okay, well we have the error and ~ the output you know ~ and the set point don't match so we have this error and so then we adjust it based on what the the three ~ PIDs are set on
and it it will it will do it so it creates a an okay one but again it we start getting into ~ you know I start saying like okay well it's it within the time domain so what happens if you know it comes up with the solution where zero is on the bottom how does it how does it deal with that? And normally it's like, ~ whoops, I forgot to do that. So let me take that back into account and then it'll make
it more ~ complex. And then it's just like, Okay, well sometimes the error is so minuscule that we don't need to change it. you know, you need to put endpoints on there. Well, ~ when it's doing that it doesn't really know how small the error needs to In oil and gas industry, we're not that critical for ~ decimal points, but in like pharmaceutical and I am assuming ~ also food industry is a lot more niche or a lot more advanced in that particular aspect
for temperature sensors and ~ different things like that. So again being able to do those is an important aspect of it and then also ~ just even knowing the error from what you can get ~ from those so again if you get something that is you know, say for example, ~ I'm asking okay, how much you know how much steam is gonna be ~ pumped out of a steam generator. you know, steam can only be between zero, which is water, and a hundred percent,
which is just ~ pure steam and ~ or superheated steam. So again, if it pumps out a hundred and ten percent, that's kind of a weird thing ~ in order to do. But another great, you know, kind of base case with that is as you said, Turbines and compressors operate a little bit differently based on their environment. So if you actually have a a turbine in a very cold and it brings in that ambient air, you can actually have a higher efficiency
than 100% on a gas turbine when you're solely looking at that. So it's just being able to know those details of like, ~ you know, this is getting that out. I can tell you why. whereas a lot of times when an AI spits out information, it can't really tell you the why behind it or why we're doing things the way that we're doing it. So it just depends on that. And that goes to again one of the things we do a lot of is compressor control.
And one of the things that matters quite a lot is where the sensors are and what sensors we actually have in the system. Because ~ for some people we We calculate ~ what you know the surge line and we calculate the performance differently, whether it has the ~ pressure in suction, whether it has the pressure in discharge, whether it has a flow, whether we're calculating ~ the amount of flow from ~ the amount of power that the ~ motor is using or the driver behind it.
So again, there's just a lot of different ways that we can set it up and a lot of different ways that we can calculate things. And it's just being able to kind of know a lot of the nuances in being able to do that. And especially when you start getting into you know, when something's running, normally it's pretty good, it's hopefully relatively steady state. But especially in the start up and shut down, it's really hard to control those factors or know what's throwing off
the start up and shut down ~ as well. So it just depends. And I think that that's one of the things that's important for companies to know is to know the process behind what they're doing and how they're doing it. And when AI spits out information about it, not to trust it blindly, but make sure, okay, how is it getting to that answer is a very important aspect. another thing that I do a lot with with new AI when they release new models
of it is we have a very simple program where I can put in a compressor map and I can ~ put in the inlet pressure, the flow rate, and the inlet temperature and it can spit out wherever it is wherever it falls on the compressor map and wherever in the outlet conditions of it essentially. And one of the more interesting things about that is I ask AI to do that. and it still kind of has a lot of issues with numerical methods. So ~ it wasn't
This last one, but it was the previous one where normally what you do is you just create a you know, you spline a curve ~ on there with a compressor map, and then you can read it anywhere on that compressor map, and it should be able to calculate that. And it looks like what I was doing in AI or what AI was feeding back to me was it was just randomly guessing what my RPMs were based on the compressor map and just randomly
guessing a spot on there instead of creating that function that you can use to calculate based off of all the inputs. And so, and then when I try and tell it, hey, create this function that you can create in Excel pretty easily, it was struggling to create a function that was working correctly in there. So I'm curious to see how I and I'm sure some people probably have different examples of this, but I'm curious to see how it will progress
of being able to create functions like that, being able to derive ~ more things that especially with a lot of people who are ~ numerical numerical method inclined. that ~ they will fix in that. So it's as I said, it's a pretty interesting ~ aspect for AI to improve upon in the future. I think you brought up something that's super important here, which is that when it comes to a person, if someone just because someone has an A in English doesn't mean you're having
an A in math, but we assume that about AIs. And it's Probably the biggest assumption people make, it's such a different skill set. There's people that are really good at math that don't know how read because so different, and vice versa, of course. So I think that's a really great point to bring up, which is that it the further we move something away from its area of expertise, including me, and include it's just like any other tool, the less skillful it gets at it.
And I think you brought up something really important. Now I love. The precision you've talked about and the kind of the lessons you've shared with us today, John. This has been really, really exciting. For people that need precision engineering and are very interested in what you're doing, kind of want to know the other areas you work in besides compressors and turbines and especially about your physical modeling, all that's amazing. Where's the best place to see what you're doing online, see your company's website and all of
those things? Yeah, sure. I work with Statistics and Control Inc. we're at www.stctrl dot com. It's one of the hardest things to say on the ~ phone and on the web. And then ~ you can also check out my LinkedIn as well. We tend to post a lot of that and then my company has a LinkedIn. So if you just s search for Statistics and Control Inc, we're based out of West Amoine, Iowa and we're pretty easy to get a hold of.
So ~ if you ever want to contact me, just feel free to add me on LinkedIn and I normally respond pretty quickly on there as well. Amazing. I'll make sure to put that in the show notes and below the video on YouTube. Thank you so much for being here for another amazing episode of the Artificial Intelligence Podcast. Thank you.
More on ai in manufacturing, energy and infrastructure
