VERN STEFANIC: Hi, and welcome to AAPG's Energy Insights. I'm Vern Stefanic,
and this is another podcast from AAPG in our ongoing series, Digging Deeper,
featuring conversations with the speakers for this year's AAPG, AAPG
Foundation's Distinguished Lecture series. Our guest today is Dr. Michael
Pyrcz, an associate professor in the Cockrell School of Engineering at the
University of Texas at Austin. Incidentally, Michael's DL lecture-- which is
available for either downloading or streaming on the AAPG website-- is titled
"Data Analytics and Machine Learning for Energy Geoscience and
Engineering." Talk about timely, right?
And today, we get to know a little more about Michael-- his journey from
rural Canada to the corporate world of energy to the position he holds today,
which continues to be both connected to and also impacting and influencing the
energy industry. And we'll talk about machine learning.
Michael, welcome to Digging Deeper. We're glad to have you with us today.
MICHAEL PYRCZ: Thank you very much for having me. It's a pleasure to be here
VERN STEFANIC: Yeah. And so your talk, as I just mentioned, is about machine
learning. And let's get right into it, because you have what we've found to be
a pretty cool life-- a pretty cool journey in your life. Where you started out,
how you went through the industry, and now at an academic institution, but
still talking to industry. And a lot of what you talk about, though, deals with
machine learning, which you have sort of been part of for many years. But it
wasn't really called machine learning, is that right?
MICHAEL PYRCZ: So, clearly, it's a case of we're all learning together right
now. There's a wave of the technology going through. But the way I see it is
machine learning is statistical learning. And we have been working with big
data. We've been working with statistical models of the subsurface, combining
multiple sources of information, working with uncertainty, and putting that all
together for the purpose of inference and prediction the whole time. And so,
for me, it's very natural for us to make the extension into what would be more
properly known as machine learning nowadays.
VERN STEFANIC: OK.
MICHAEL PYRCZ: So, yeah, I would say, I think I've been working in that area
for a while. And I think many of us have been working in data analytics and
VERN STEFANIC: So how come everybody's scared to death? Not everybody, but a
lot of people. You hear about-- there's an intimidation factor about the
concept of machine learning. What's going on?
MICHAEL PYRCZ: So I think there's a lot of misunderstanding about the
technology. Now, I think it can be-- it's interesting, there's a whole
continuum. It can either be quite incremental-- in the form of automation of
some of our common tasks-- and in some cases, it may be disruptive or
transformative. I like to say transformative and be more positive about it. But
no matter how I see it, I think there's great opportunities to see new value,
to be able to make our jobs more focused. Geoscientists will spend more time
doing geoscience, that's how I see it. We'll spend less time clicking in our
interpretations. We'll have more guidance.
I think sometimes our interpretations are very challenged by all of the
different types of data that we try to integrate. And what we'll find is that
these new tools will help us avoid contradictions. I've seen that before,
working on subsurface asset teams, where I don't think everybody quite agrees
around the table. And then it's not until you build the model that you see
there's contradictions in the concepts, right?
VERN STEFANIC: Yeah, yeah, yeah, yeah. So in talking and giving all of us an
idea of the scope or the magnitude of what's going on with machine learning, is
this something that industry-wide everybody's trying to pick up on right now?
Or is it just a couple people are saying, we're ahead of the curve?
MICHAEL PYRCZ: Definitely different companies are at different positions as
far as readiness in this digital revolution, that's for sure. But just last
year alone, I think I visited and taught within about 20 different companies. I
do still teach my courses, in case the Chair is watching this video, right?
Now, but I'll tell you what-- every single company I go to, they're facing the
challenge and they're trying to do something right now around digital
technologies, data analytics, and machine learning. And I've seen a whole range
of responses. I've seen a lot of companies who are-- we have a couple of
people, maybe they've hired a couple data scientists, we're trying to fast
follow or trying to understand what's going on. I've seen other companies that
say, we want to be at the lead.
There's one company-- Conoco Phillips-- they have a huge effort to teach
hundreds of engineers and geoscientists to make them citizen data scientists,
which amazing effort, amazing program. And I've seen a whole range in between
there of different efforts around this. I have seen, as far as adding value, a
lot more focus on the data. And to be honest, you know we have P10 and P90?
VERN STEFANIC: Mm-hmm.
MICHAEL PYRCZ: To me, the P10, or the low side outcome-- a Chevron P10--
would be, we just do better with data. We do better with understanding
statistics. And more people understand how to do basic coding and scripting so
they can automate and be more efficient in what they do. I think that-- to me,
that might be where we see much of our value.
VERN STEFANIC: So what is the state of the art of data, right now, that the
industry is using?
MICHAEL PYRCZ: So it's a very good question. When I go from company to
company, I see similar types of challenges. Data storage and curation is a huge
challenge. What I'm seeing is a lot of data sets for which there is a
combinatorial of versions because of multiple interpretations, uncertainty,
data that needs to be integrated into the-- or concepts have to be integrated
together. And the result is, often this is not stored in such a way that, as we
move through project life, which often has, as you know, cycles of many professionals
working on the same project, right? People don't stay on the same project.
Things can fall through the cracks.
VERN STEFANIC: Yeah.
MICHAEL PYRCZ: And we may not fully harness what was done previously. I see
there's many issues around data like that. I think when it comes to metadata,
we also do struggle as an industry. How do we track-- how do we keep track of
all of the assumptions, all of the choices that went into formulating that
data? Many of the data we work with isn't really primary data, right? It's not
directly from the measure, it required layers of interpretation on top.
So I think-- to be honest, I think, about a decade ago, I saw a lot of
companies go through a revolution, saying, we're going to ramp up and have this
great universal database. And they start to look at database technologies,
which in itself is its own scientific field. And they were looking at, how do
we have standards, how do we put it together, how do we get everything to talk
to each other? I don't think we succeeded there. I don't think that finished
for most companies. And so I think we're still going to face challenges with
data as we go into data analytics.
VERN STEFANIC: So most of the challenges now-- we may not even know what the
big challenges in interpretation are going to be, because we still don't have--
it's garbage in, garbage out, so to speak.
MICHAEL PYRCZ: Yeah. And so this is part of the reason I'm motivated to
teach. So I've worked to actually completely change the curriculum for our
undergrad program at University of Texas at Austin. I'm doing that in my
department, I'm also helping over in the College of Natural Sciences with the
freshman research initiative. And I want scientists and engineers to understand
statistics, programming, data analytics, and machine learning, so that as they
go out there, they're able to pose the questions.
I had a student come in my office and they said, I've got this great idea--
I'm going to take seismic information, well information, I'm going to do these
great forecasts. And I said to them, I said, what are your predictor features?
And they didn't know. They hadn't broken it down to the nuts and bolts. They
didn't understand well enough to be able to pose the question in a manner that
could be used by data analytics and machine learning.
VERN STEFANIC: Actually, I want to come back to some more of your
experiences in the classroom and teaching. But getting there, the one thing
that is obvious in your lecture, and is apparent even from us talking here
right now, is that you have a passion for learning about things, seeing
challenges, and figuring out how to go forward through them, and trying to take
steps forward, always. It seems to me like everything is advancing, advancing.
Could you talk a little bit about your background? How did all that start? For
some of us who don't know you that well, where are you from? What's going on?
MICHAEL PYRCZ: So I think I can best describe myself as a farm kid from
Alberta. I came from a small town. I didn't grow up on a farm but I was working
on a farm. I spent a while during high school working as a waiter at a local
restaurant. And then went into a dairy farm and realized I work with cows
better than customers in the restaurant. I really enjoyed the cows and I really
enjoyed the manual labor and all of that aspect of it.
OK. So I came out of that. And from a family that actually was, I guess, we
could characterize as low income, not a lot of opportunities. I'm first
generation to-- I'm the first to get a university degree from the family. So it
was kind of a big step. I went out and just did that. I was driven to do that.
And what's really funny is I hadn't really thought about university, I had no
concept of it. And it was just a random conversation with a university student
that led to the whole thing. I was in grade 11, I was working full-time, had a
girlfriend, had the car payments, had all that stuff, and I was doing terrible
in high school. I wasn't very good at all, my marks were terrible.
VERN STEFANIC: You were not a good student.
MICHAEL PYRCZ: I was not a great student at that point. So grade 11--
fortunately, this was Canada. I'll get back to that. So I'm filling up my old,
rusty car, and another person pulls up at the pump on the other side on a cold
evening-- dark and cold evening there in Alberta. And he starts talking to me.
And this is a Canadian thing, we talk to strangers. We totally do. And so he
looks at me and he says, do you know how that engine works? And I said, what do
you mean? I knew something about the internal combustion engine, four cycle,
whatever it is-- four stroke. And he said, no. He started to explain the
theoretical carnot cycle, about entropy, and pressure, and temperature. And
they were an engineering student working materials engineering, and talking about
how they were building--
VERN STEFANIC: You didn't know this person?
MICHAEL PYRCZ: Didn't know the person. And they drew the chart, the entropy
plot, directly on a windshield with frost. They were telling me these things.
And it was amazing. I was just taken aback. I'd never really known engineers or
geoscientists or scientists, and they were explaining these scientific
concepts. And I looked at that and I said-- that night, I said, I'm going to be
a scientist. I'm going to be an engineer. I'm going to learn this stuff.
So the next day, I went to high school. I made an appointment, met with the
student counselor, Mr. Jamison. I've thanked him later on, don't worry. And Mr.
Jamison, he looked up my marks, he looked at me, and he said, Michael, university
is not for everybody. University's not for everybody. My marks were not even
close. And from that moment, I cut down the hours at the restaurant, I focused,
and that was it. That was my mission, to go into university.
Well, I got accepted in engineering. Part of the drive for that was the
engineering program-- they had a mining engineering program that was very
geologic, but it also had a co-op program, so I could work half the time. The
geosciences didn't have that. So for my undergrad, unfortunately, I was forced
to do mining engineering instead of geoscience, which is what I really wanted
to do. And I graduate number one in my class. I was on fire, I was driven. I
was just like, this is it, this is my life now. And so I went all the way
through to PhD, which was much more geologic, subsurface, modeling, and
VERN STEFANIC: Where was the education? Where did you go to school?
MICHAEL PYRCZ: Both degrees, University of Alberta.
VERN STEFANIC: OK.
MICHAEL PYRCZ: Up in Edmonton, Alberta. So that explains things-- the cold,
frosty evenings in the winter. I think you could understand how bad it could
be. Minus 40 is not atypical.
VERN STEFANIC: Wait, did-- probably not, but I got to ask. So the person who
you had the conversation with-- did you ever-- who was it? Was it-- did you
ever find them?
MICHAEL PYRCZ: So I know that they were a student-- kind of a senior
student-- from the University of Alberta in materials engineering. But I never
found them. I thanked my guidance counselor. Because Mr. Jamison, he made me
VERN STEFANIC: Yeah.
MICHAEL PYRCZ: He made me so mad that day. I said, what do you mean, it's
not for me?
VERN STEFANIC: Yeah.
MICHAEL PYRCZ: I can do it. And it motivated me. But, no, I'd love to meet
that student again. But I'll tell you what-- now I do the same thing. To me,
the whole world is a gas station. I'm serious. I walked out of campus, as I
often do, like 8:00 at night because I'm so busy. And I'd forgotten to eat all
day, because I'm just way too busy as a professor. And I walk out and there's a
student just standing at the back door in the dark there, propping the door
open. And I thought, well, as a professor-- I thought it was a student-- I
said, as a professor, I should ask, what are you doing here, kind of thing. Is
there a security issue or something, the door being held open at night and all,
right? And they said, well, there was some type of event inside they were
attending. It was a Lan party, actually. A bunch of the engineering students
that got together and playing Lan games, or whatever.
And then I said-- so I looked at them and said, are you a student here? And
they said, no. And then I looked back and I introduced myself, I'm Professor
Pyrcz. You're very much welcome here, nice to see you. They introduced
themselves very politely. And then I said, you should hang around here a little
bit more often, we'll make you into a scientist or an engineer. And I just
thought what that meant. Like if it had been me in their shoes, for a professor
on campus to take the time to welcome me and to speak to me like that--
VERN STEFANIC: Yeah.
MICHAEL PYRCZ: Would have potentially changed my life. I thought that way.
And so every time I meet with people now-- I'm funny. I go to the HEB store, I
tell people I'm a professor, because it always ends up in a conversation. If a
bunch of students are working and, OK, what are you going, where are you going
VERN STEFANIC: Yeah. So there's a part of you not only that wants to learn
about more, but you also have the other side wanting to give back.
MICHAEL PYRCZ: You want to share it, yeah.
VERN STEFANIC: Does that come from your family? Does that come from the Boy
Scouts? What was it?
MICHAEL PYRCZ: So it's a really good question. I grew-- I think my situation
growing up was pretty difficult and very challenged. And I think, because of
those struggles, I have a high degree of sympathy. I can put myself in other
people's shoes, I can imagine. And so I think that partly motivates me. I think
the professor I studied with during my PhD, Clayton Deutsch, he had developed
open source software for geostatistics that was used all over the world. It was
used to develop all kinds of software packages that made-- I'm sure-- fortunes,
and he didn't worry about it. He just freely gave things away. And what I
learned from him was when you do that, it does come back.
And for me, what am I concerned about? I'm concerned about finding funding
for students, helping my students do well. And so what I've done now is I give
every one of my lectures away. Now, part of it is also just practical. In this
modern digital age, if you record your lectures and give it to students, guess
what? They're going to give it away. Do you know what I mean? There's internet
sites where everything can-- my book that I wrote with Oxford, you can get it
on some site right now, the PDF.
VERN STEFANIC: Right.
MICHAEL PYRCZ: So what I realized was, just give it away myself. And so I
record every lecture, I record a product that's useful for professionals, useful
for my students. People appreciate it.
VERN STEFANIC: Yeah.
MICHAEL PYRCZ: And I get to be part of something, I get to do something
great. And the result is-- I'll tell you what-- I go into meetings with
companies and I'm like, shall we talk about what I do and then we can talk
about maybe collaboration? And they said, don't worry, we follow you on
Twitter, we watch your lectures on YouTube, we know what you do. Let's talk
about collaboration, let's talk about funding students. That's a dream as a professor.
To me, that works perfectly.
VERN STEFANIC: Way cool, way cool. If you could fill in a blank or two
before we-- we keep wanting to-- I keep wanting to get to the education and
some of the ways that you interact with students today and everything. But even
before that, to fill in a couple of blanks. Because you were in industry for a
while. So you get your PhD.
MICHAEL PYRCZ: Yes.
VERN STEFANIC: And what happens career-wise after that?
MICHAEL PYRCZ: Straight to Chevron.
VERN STEFANIC: Straight to Chevron.
MICHAEL PYRCZ: Working in a technology company, yes.
VERN STEFANIC: OK, OK. You're doing what?
MICHAEL PYRCZ: So starting out, it was more practitioner, working on service
projects, I'm getting some experience doing subsurface modeling in the
industry. During my PhD, I had supported myself and my PhD by consulting at the
same time. So I'd done a bunch of consulting for a bunch of energy companies
out of Calgary. And so, when I was in Chevron, I got--
VERN STEFANIC: You say that casually. That's pretty impressive.
MICHAEL PYRCZ: It was good. It was good. I joke, I took a pay cut when I
start working with Chevron, out of my PhD. I had two kids while I was in my
PhD. And my wife was home with the kids, and I was providing for a family while
doing a PhD. It was great, it worked. And I was fortunate, I had great
But when I started with Chevron, I had an opportunity to learn the Chevron
way, the Chevron methods, to just drown in amazing data sets. And to be
mentored-- you know, you get a PhD but, come on, you've got to realize-- the
people in the industry have been doing it for 30 years, sit down and just learn
what they know. You know what I mean? So people like Scott Meadow, Sebastian
Strebelle, who was my team leader, Morgan Sullivan-- amazing mentors. I got to
work with Henry Posamentier for a while.
VERN STEFANIC: Wow.
MICHAEL PYRCZ: Just great people. Mitch Harris. I just love these people.
VERN STEFANIC: Great names, great names.
MICHAEL PYRCZ: It was a very wonderful-- it was a wonderful time to be
there. And so I got to learn a lot, grow up in that. And then I found that I
really liked research. And so I start to develop, I start to code, I start to
learn more quantitative methods, the data analytics, more of that. And it
reached a point where I start to lead the team, lead the projects, lead the
program. It was several million dollars worth of research a year that I was
VERN STEFANIC: Yeah.
MICHAEL PYRCZ: And then, right around that, the opportunity to go back to
academia appeared. I'd written a book, I'd had about 40-something peer-reviewed
publications while still at Chevron. And academia came knocking.
VERN STEFANIC: By the way, Henry and Mitch Harris, oh, wow. Great, great,
MICHAEL PYRCZ: I love them.
VERN STEFANIC: Industry-wise, profession-wise, AAPG-wise. But through this--
before we leave-- since you were doing research, I mean, that that would be
perfect for the way you are in wanting to keep learning more and learning more
and learning more. Was this in an era-- and I know we've talked before that
machine learning has always been with us, we just call it different things--
but is this where some of those real ties and connections started for you?
MICHAEL PYRCZ: I think-- thinking about the world and our data sets and what
were the new lenses we could use to try to explore and understand that. It was
a general mission of, we can quantify everything. It's kind of been my attitude
the whole time. And if you think about it, that's the prerequisite for
everything machine learning. You have to feed in numbers, you can't just feed
in some subjective interpretation. You have to quantify. And so I think that
really expanded my interest around data analytics.
And then, of course, we got into basic machine learning methodologies-- all
over the place-- but we call it more statistical learning. Machine learning did
take off just in the last couple years. I know this is the fourth wave of it,
but as far as taking off again. And so now we've-- now, at the same time, when
I teach my machine learning course, I say something. And I say, if you take my
data analytics, my geostatistics course, I can stand up here and say, I wrote
the book in it. I know it, I've published lots on it, I'm an expert, recognized
expert. What I say in machine learning is I speak from humility. And I say,
guess what, we're all learning together a little bit. Because the field is
moving so fast. There's so many things. But I'll tell you what, understanding
fundamental statistics and data analytics, you can understand in depth what's
going on with machine learning, generally.
VERN STEFANIC: OK. Was there a point-- I want to say this politely and
without sounding snarky-- when you realized corporate culture wasn't for you?
Or was that part of the realization that you wanted to give back?
MICHAEL PYRCZ: So I'm kind of-- I'm unusual. Because I got to tell you,
there's many days I wake up and I still think, what did I do? Because I love
industry. I love my experience within Chevron. I tell my students, in my
classes, I say, I wish my career upon you. Because I was so fortunate, I loved
my experience inside the industry. In fact, I mimic industry still. I have many
students who want to study with me not just because of the topic, because of my
behaviors. I run my team in the university like I ran my team over in industry.
So the same attitudes around leadership, transparency, respect, and so forth. I
use that as a professor.
And I think people-- and you know what I think? I think if we use leadership
skills as professors, won't we send out people who will be better leaders?
Shouldn't we be examples of that? So anyway, what I'm trying to say here is
that I love industry. In fact, I remember the moment. So I interviewed. And
then, of course, I went through a process just like everybody did. There was
interest in me, but I went through the fair interview system. And they-- I knew
they were coming back with an offer. And I had a number on a piece of paper and
that number was the minimum I could accept-- because, of course, it was a pay
cut going to academia-- and I could still take care of my family. That matters
to me, I've got three kids.
VERN STEFANIC: Yeah.
MICHAEL PYRCZ: The Chair of the department sent that number. And I actually,
probably, made an audible, ugh, darn. Because there was a part of me that
wanted them to be under, and then I would just have to be, sorry, I'll just
have to keep being happy in industry. But no, they were right there at that
number. And it was my wife who sat down with me afterwards, and said-- and I
start to rationalize, I said, but no, we can't, it's a big move, it's all this.
And she said, you will always regret, you will always wonder, what if? And so I
made the choice.
So what I'm saying is, there is a parallel universe in which I'm still happy
in industry. And I came to academia really driven by-- so when I was in
industry, my attitude was this-- accept every challenge. And I did. I did a lot
of interesting things because I said, I'll learn that. Like I learned C++ to
take an internship with Chevron right at the beginning. I had done nothing in
C++, and within a month I could do reasonable C++, that kind of thing, right?
VERN STEFANIC: Yeah.
MICHAEL PYRCZ: So to me, this was the ultimate accept every challenge. Could
I be a professor?
VERN STEFANIC: Well, that just begs the question-- to jump out of sequence
here, though. So what has been the biggest challenge that just kicked your
MICHAEL PYRCZ: I know. I know, I know. So I like the culture inside of
industry. We win together. It's taken them two years to teach me that the Chair
is not my boss. I still believe-- I feel like I'm in a hierarchical
organization. I feel like everyone around me is on a team together. And my
attitude-- I have 12 PhDs, six of them are co-supervised. And I'm more than
happy to do that. What an honor to work with Dr. Torres-Verdin and Dr. Lake and
just so many great professors, right?
OK. So academia-- I'm not saying anything negative, just to be clear here--
but that same focus on team, that same unified-- you in academia will feel
alone. I'll tell you, I went to do some teaching at Noble Energy. And the
second day of teaching, I went-- I woke up that morning, I started my car, and
the car wouldn't start. And I realized, I'm just a professor here by myself
with my personal car, it doesn't start. How the heck am I going to get there?
How am I going to teach? How am I going to get my car fixed while I'm teaching
all day? And that's when I realized, if I'd been working with Chevron, there's
a card in my pocket, I'd phone it, everything's taken-- the company's got your
VERN STEFANIC: Yeah.
MICHAEL PYRCZ: You don't have that in academia. And so everything you do,
you've got to be entrepreneurial. You've got to find funding.
VERN STEFANIC: Right.
MICHAEL PYRCZ: And people don't understand, one PhD, without overhead, is
about $50,000 per year. And there are companies, many companies, that will
nickel and dime you about $10,000 and $20,000 a year. And you're sitting there
in the back of your mind realizing, this is just covering a term. You know what
I'm saying? So I find that, to me, those are some of the greatest challenges.
Now, are we going to ask about the positive stuff so I can end on a positive
VERN STEFANIC: Certainly. I was-- that's exactly what I was wondering. Yeah.
What about the positives?
MICHAEL PYRCZ: The positives. So the days that I came home electrified, on
fire, when I was in industry were the days that I was teaching and mentoring.
My life is full of that now. I have an open door policy, students are coming
in, I'm constantly mentoring and helping and assisting and teaching. I love
teaching. I really enjoy standing up in front of a classroom and just teaching.
I'll teach a new concept, like bootstrap is a good example. And I'll say, OK,
today we're going to learn about statistical bootstrap, who's heard of this
before? Nobody puts their hands up. And I'll look out and I'll say, this is
truly a great day. I get to be the one to share this with you. I love that,
that just gets me so excited, because this is changing people's worlds. And I
have so many students-- can I brag?
VERN STEFANIC: Yeah. Oh, please.
MICHAEL PYRCZ: My rankings are really, really high. The rankings are very
high, and the students love it. And I got so many students sending me messages
of, this was the best class they took and they're so excited by it. And it
really moves me. And to me, that just keeps me going. Because it doesn't matter
what's difficult, when I get that, it just makes me feel so excited about what
I'm doing. I feel validated about this choice.
VERN STEFANIC: I'm curious, because you do teach classes in industry as
well, the difference-- two things-- the difference in your approach, if there
is a difference, but also the difference in the people who you're talking to.
MICHAEL PYRCZ: Yeah. Yeah. That's a very good question. So one thing I do
is-- well, when I teach in industry, I do recognize the fact I don't get as
much time with the people. And so I high grade a lot more. And I do tend to be
much more just practical. My attitude when I teach in industry is I want to
teach you something in an hour or two that you can use tomorrow to show
incremental value. That's my goal. And so I do a lot of that. I may cover some
basic fundamentals and theory, but then I have to switch into practice pretty
With the students, you can evolve. You can grow that over time. And I do
enjoy having a-- teaching a full term is-- it's amazing what you can cover in a
full term, right? Now, I'll tell you one thing, though. I am a very nice
person, but when it comes to being a lecturer and control of the classroom, I'm
awful. I'm really awful. I will not tolerate unprofessional anything. In other
words, your cell phones are away, if you've got a laptop-- and I roam that
classroom-- and it's not on the course content, I'll kick you out. Because I
have no tolerance for that, because guess what? You're going to leave my classroom
knowing professional behavior.
And when I was a team leader, if you're sitting there texting someone in the
middle of my meeting, it looks unprofessional. And so I use that when I teach,
I have the professional industry standard. And I'm finding that's not
universal. Students are surprised that, oh, I can't have my cell phone out, I
can't be texting someone while I'm in your class?
VERN STEFANIC: Yeah.
MICHAEL PYRCZ: It amazes me. So anyway, it's-- yeah. I take the culture from
industry into my classroom.
VERN STEFANIC: And there's so much-- I was thinking while you were telling
that story. So I taught-- nothing like geoscience-- script writing for a while
at a college in Tulsa. And I remember, maybe it was five, six years ago, all of
a sudden-- and at first, I was really hacked off when they had their computers
and everything. And then I realized that was their tool, they were actually
using it. But they better be using it as a tool or else you get mad.
So machine learning-- in your analysis, then, is it a tool or is it the
magic solution to something? How should we think of it?
MICHAEL PYRCZ: So that's a really good question. People need to manage their
expectations with the technology. So one course I've really enjoyed to teach is
I have a half a day course that I teach to executives in industry. It's
basically-- I'm not going to say it's buzzword compliance-- but it's teaching
them the highlights-- how would they judge this technology, how should they see
the technology, what are some of the terms being used. And so I really enjoy
teaching that. Because what I do is I spend about an hour just being very
critical, very negative about it, showing how things can crash, how things
don't work, when it could be abused and misused.
Now, what you'll find about me is I'm pretty flexible. I have a whole
continuum of beliefs depending on the application. On one side, I'm quite
optimistic that there's huge value and opportunity just to dramatically or make
a significant shift or disruption to the way that we get the job done for the
subsurface and geosciences and engineering. On the other side, I'm kind of
pessimistic. And I see the limitations, I see how we're not going to be able to
completely automate everything. And so, in general, my response would be, I'm
never an advocate for the geoscientists in the box. I don't think we ever do. I
go back to the idea of these methodologies will allow us to have geoscientists
spend more time doing geoscience. And some of the more mundane parts of their
job, their tasks, will basically be automated. Their interpretations would be
better supported by systems that help them semi-automate their interpretations,
help them put their information in, their knowledge in, at the same time, be
aware of all the other potential contradictions with other information sources.
So in general, that's been my perspective as far as how these technologies
are going to change and move us forward.
VERN STEFANIC: Yeah, because the deal-- I remember in your lecture you
talked about it's an enormously high percentage of time that we're actually
worrying about the data itself, right? What was it, like 80%?
MICHAEL PYRCZ: Often could be 90% of data preparation. Yes.
VERN STEFANIC: Yeah, right.
MICHAEL PYRCZ: Yes. And so, clearly, there's a lot we can do to automate and
improve the efficiencies of what we do. And so I think there's huge value in
that. At the same time, what's very interesting-- I was an a PhD defense, and I
was co-supervising the student with Dr. Torres-Verdin. I think it was one of
our co-supervised students. And I asked something about data analytics and
machine learning, and he stopped and he said, listen, sometimes I feel like
people use it as a crutch. And there's a great quote from Dr. Torres-Verdin
where he said, people, instead of understanding the fundamental geoscience and
engineering, will try to use data-driven approaches as a crutch.
That's my fear is that that will drive us in that direction. I don't think
it should be that way. It's like I said during the talk, we have the first, the
second and third scientific paradigms are essential to what we do. Geoscience
and engineering knowledge is central to what we do. The fourth paradigm and
data-driven approaches provide us a new toolkit, new methodologies we can use. Interpretability
What I find with my students is they often jump to the most complicated
models. In my classes, they'll get so excited, they want to do the really
complicated methodologies. They're like, look at me, I did this deep
convolutional generative adversarial network, blah, blah, blah. And I built
this model that made this prediction, it's got this r squared of variance
explained of 90-something percent. And I'll go back and I'll look at it and
I'll realize that this is over fit. This is just-- yeah, it's a model that has
so many parameters in it and so little data to train it, you're just able to
precisely honor the data. It has enough degrees of freedom to honor the data.
And so this has been one of my concerns, too. Now, at the same time, I am
excited by the technology. I show an example in my talk where we're able to
make forecasts of the intragranular velocity field of flow of a fluid, right?
That's awesome. Now, where do I think that could play a role-- multi-scale
modeling. Imagine we're trying to understand how poor scale cementation,
diagenetic alterations affect overall recovery mechanisms at the large scale of
the drainage radius of the well. Imagine if we had models that can work that
fast in making those predictions across those small scales to larger scales.
That would be beautiful. Because now we have the geoscientists sitting there
saying, I think this is the diagenetic alterations, I think this is how the
grain structure is stacked, I think this is the uncertainty range. And they can
immediately see how that impacts the bottom line, which is dollars and flow,
VERN STEFANIC: Right. Right. OK. So let me ask you this question, then, with
that in mind. So how is this going to impact the industry of the future?
MICHAEL PYRCZ: Yeah. So we had a great talk-- one of the Vice Presidents of
Ecuador came and visited us at the department and gave an excellent talk about,
basically, where Ecuador is going. And he made a lot of statements about how
this new technology is going to be used, how the workforce is going to change.
I think there will be changes. Now, my perspective is more of an evolution
rather than a revolution. I was in front of a panel for Price Waterhouse
Cooper, as I mentioned. And there was another individual expert there who was
suggesting a massive replacement of engineers and geoscientists with data
scientists. And I look at that-- or they are advocates of the geoscientists or
the engineer in the box.
VERN STEFANIC: Which is, I think, where some of the fear and intimidation is
coming from, right?
MICHAEL PYRCZ: And I look at that and I don't agree with that at all. I
think looking forward-- first of all, I look at our business, and I know our
business because I've worked in it for all these years. I know much of it, and
I know the complexity. I have a respect for the scientific disciplines. I've
sat and listened to Professor Pemberton talk about ichnofacies. I've listened
to Henry Posamentier talk about how he can interpret and do analysis of seismic
and seismic geomorphology. I've listened to Tim McHargue talking about sequence
stratigraphy standing at the rocks in the Karoo Basin in South Africa. Morgan
Sullivan in the Ross Formation-- and so on and on.
I know the complexity of what we do. And this is what I know-- the great
innovations and value that's been added within our field is because of
creativity. It's because of discovering new concepts. And no machine has ever
done that. There's a perfect quote I have in one of my talks-- my lectures to
executives-- where it talks about the fact that the Higgs Boson was discovered
using a lot of data analytics and a lot of machine learning and a lot of data
science behind it to support it. But the actual discovery was not data-driven.
It was not based on those sciences. They support it. But the actual creativity
to look for it, the actual development of the design, it didn't come from a
So that's what I believe. I believe that we'll still need-- geoscience and
engineering will be core to everything we do.
VERN STEFANIC: I want to ask a personal question in that light, then. That
would be-- and this gets us back to the story you talked about when you were at
the-- filling up your car in Canada, right? And I think there may be a
connection with what you just said and that moment and this long arc you've
been on, this journey, is your creativity and your creative approach to life.
Where do you go for inspiration? And where does that creative drive come?
MICHAEL PYRCZ: So it's a very good question. I feel I'm in my element when
I'm working with my hands, when I'm doing something difficult physically. And
so I have kayaks-- I actually have eight kayaks-- in my garage. Come to Austin,
come kayaking with me. I take people kayaking every weekend. I tell my students
or other faculty, I'll be 8:00 in the morning, Saturday, show up at my house,
come kayaking. And I take people all the time.
When we do a really hard kayaking trip, when you have physical exertion,
when you're paddling into the wind and the waves are battering against you-- to
me, that's beautiful. And, to me, that teaches me something and it makes me
appreciate something. Long hikes, climbing things, mountain biking things,
doing things on my mountain bike I didn't think I could do. My wife and I, we
do some weight training. I know I'm a skinny guy, but we still do a little bit
to stay healthy and strong-- keep our bones strong. When you're lifting
something heavier than you thought you could-- to me, there's beauty in that.
And so sometimes when I talk to my students-- my graduate class, my very
first graduate class, was pretty small. It didn't sell out. I was a brand new
professor, nobody knew me that first fall. There was only six students in it. I
took everyone kayaking. I love it. I love it. I think that's-- I think part of
education and sharing is sharing some of these things. I also do talk to
students about work-life balance. I talk to them about how to succeed in
industry. Those are other topics. But I do find that that's-- and then, of course,
besides that, it's in those moments you pick up a guitar and just strum
something and play away at something. It's just the feeling of the vibration
and hearing the music you create, or attending music. So much beauty.
VERN STEFANIC: There's much beauty. And you see much beauty, so this ought
to be an easy one to end with, and that's looking forward. And what
possibilities do we still have? Have you tried to-- and you've said it
yourself, that everything is changing so quickly and we're all learning together
at the same time. So this may be difficult. On the other hand, you're a pretty
optimistic person. What do you see going forward? What kind of expect-- what
kind of changes might we anticipate?
MICHAEL PYRCZ: Yeah. Good question, good question. So there was a great talk
given by my old PhD advisor on black boxes. And the comment was that black box
use is not a bad thing. How many people truly understand their cell phone?
Because the cell phone rate, the technology behind the cell phone, none of us really
understand it-- many of us don't. But we use it, and we make use of it, and
we're productive with it. The idea of the maturing of some of these
technologies to the point where they become black box and extremely useful--
that, to me, excites me. The idea of huge, new ways that we can interface with
data and with systems and understand and explore systems, new lenses to be able
to see what's going on in the subsurface. That's exciting to me.
I truly believe that this technology will just make it more and more
exciting. It's a wonderful time to be a geoscientist. It's an amazing time to
see this transition. I think we're going to like what we'll see.
VERN STEFANIC: And I can't wait to be there with you on that. Thank you.
MICHAEL PYRCZ: Thank you.
VERN STEFANIC: Thank you, Michael, for being with us today.
MICHAEL PYRCZ: Appreciate it.
VERN STEFANIC: We've been talking today to Dr. Michael Pyrcz, an associate
professor in the Cockrell School of Engineering at the University of Texas at
Austin. Be sure to check out his lecture at AAPG.org. Believe me, you won't be
sorry. And watch this space for more AAPG podcasts that will cover a variety of
important subjects and intriguing people, including our ongoing Digging Deeper
series, featuring conversations with this year's Distinguished Lecturers.
All AAPG podcasts are available through the AAPG website or your favorite
podcast platform-- iTunes, Spotify, Stitcher, Google-- whatever your
preference, we're they're. The Distinguished Lecture program is a jointly
operated program by AAPG and the AAPG Foundation. And, as always, we hope
you'll take a moment soon to check out the AAPG Foundation website to discover
how you can be part of ensuring the future of geosciences. For now, thanks for