Podcast: "A.I., the Imago Dei, Mad Max, and the Public Square" with William Monroe, Ph.D.

Feb 4, 2025 | Podcast

Why the virtue of data scientists matters.

William Monroe (Ph.D), Vice President of Analytical AI, discusses AI’s multifaceted opportunities and dangers, including what it can and cannot do. Their conversation ranges from insights into some of Analytical AI’s work with various government agencies that help encourage the good, repair the broken, create the missing, and oppose evil. Along the way, they talk about the importance of understanding the image of God, AI as a therapy tool, MadMax, red-lining, free markets, and the public square’s role in policy development.

Listen to the episode

Read the episode transcript

Open the original episode on Substack for the synchronized transcript.

0:03 Greetings and welcome to another episode of the Faithful Presence podcast. I am your host, Reverend Michael Langer, founder, ordained minister, and president of Faithful Presence of Ministry in Washington, D.C., equipping servants and leaders to flourish as they participate in Christ’s mission of making all things new. We do that through providing spiritual, emotional, relational, vocational, ideological, cultural,

0:32 and evangelical formation. You put it all together and it spells service. and supporting the local church and equipping their members for Christ-centered participation in the public square. If I were going to give you the 10-second elevator pitch, we do whole life discipleship and the whole of life for the whole of the public square.

0:56 We’re glad to be back with you. It’s 2025. We’ve got a lot going on this month, including the inauguration of the 47th President of the United States. I actually was able to get myself a ticket to attend that. And being from Iowa,

1:13 I was trying to decide if I was gonna wear a coat or not wear a coat to that event because it’s gonna only be in the 20s. But apparently I just saw in the Wall Street Journal that now they’re thinking about having it indoors because not everybody in the country is as tough as people

1:29 who grew up in Iowa. So we shall see what happens there. But it is an exciting time today. I am excited to kind of get 2025 rolling with Dr. William Monroe. He is the vice president for data science or analytical AI, which is based in Birmingham, Alabama.

1:51 He has a PhD in, get this, the explainability of deep learning. William developed and manages artificial intelligence and machine learning solutions at analytical AI and has been doing that since 2020. He’s highly experienced in integrating computer algorithms with user-centric focus.

2:17 His PhD allows him to kind of be on the cutting edge of a lot of what is going on in the AI world right now. We’re gonna talk about that kind of in depth. And he was formerly the managing data and computational scientist at Alabama’s fastest supercomputer. And if that was not enough,

2:39 the other thing that William Monroe used to do is help lead worship at the very first church that I planted in Iowa City, Iowa, One Ancient Hope, where occasionally he would break out his dobro or some other homemade musical instrument that he had made that week. William, welcome to the podcast. How the heck are ye?

3:00 I’m doing well. I’m crazy. I’m crazy. You are crazy. How’s the wife and kids, man?

3:08 Crazy as well. Crazy as well. But it’s good. It’s good. We’re up to, you know, we’ve got four kids now. Yeah. That’s fun. Always an adventure. Always an adventure.

3:21 Okay, so when you and I were seeing each other regularly and hanging out, my understanding of what you did after you graduated had the way I think I remember you describing it to me as well. It’s kind of like… Google Maps for your chest and then when they need to do a surgery and go down there,

3:46 we can help them get the thing that they’re looking for with this pre-mapping information. How close or how far off am I from the reality of what your job was?

3:59 No, no, that’s pretty close. That’s pretty close. So what we did was medical imaging algorithms. My title at that job was Algorithm Scientist, which was a fun and obscure title. But what we did was take pulmonary CT images and then map out the lungs and lobes

4:25 and the airways so that if there needed to be some kind of intervention or measure of what’s going on in the chest, that that could be done by the computer instead of having to go in there and eyeball it with a bronchoscope or something like that.

4:42 Okay, a bronchoscope because that’s a thing that people use, a bronchoscope, which is not like a bronchosaurus because that would be painful, but a bronchoscope, much smaller. But they both have long necks. Ah, see? See? See how those things come together?

5:01 Yeah, everything’s coming together.

5:03 Okay. So tell me, how did you come to be interested in Data Science 1 and AI 2? How did that happen?

5:14 Yeah, so it’s an interesting story. So I went to the University of Iowa because I wanted to… be a rock star, which is not the normal reason why you go to the University of Iowa.

5:30 But it’s not a bad one. It’s not a bad one.

5:33 University of Iowa. But I had a lot of music that I wanted to make and I couldn’t get into a music college because I kind of decided a little bit too late that that was something that I wanted to do. And so my People in my life counseled me, hey, you know, go,

5:56 you can do math and science, go get an engineering degree. My uncle counseled me, said, hey, get an electrical engineering degree, because then you’ll understand how all the equipment works. And then while you’ve got an engineering degree, you’ll make a nice living while you’re trying to figure out the whole rock star thing.

6:15 And so that’s what I did. I went to the University of Iowa, got an electrical and computer engineering degree. And the track of my efforts there was all in digital signal processing, essentially, which Iowa has one of the foremost image processing groups in academia there.

6:40 And so naturally professors there sucked me into what they were doing with digital image processing. And at the time, AI wasn’t really a thing because the computer hardware couldn’t do it. So I remember taking a class at the University of Iowa, where we learned about neural networks, we implemented neural networks. And then the professor said,

7:03 but you’re never going to use this because computers don’t do this very quickly. And that was in 2010. And so then 2012, everybody realized, oh, that’s what NVIDIA GPUs are for. And then the whole world changed and AI really started taking off there. And so I learned about the basics of the algorithms at that time,

7:31 but then after moving here to Alabama, getting my PhD at UAB is where I became more involved directly with, the ai piece as well because it became computationally feasible to do ai once you

7:49 start using gpus okay so i i understood most of that okay um so one of the reasons that i wanted to have you on the podcast one is because i love you a ton and like enjoy your family and you are a really gifted musician, but you also are, as a Christian person,

8:15 you’re in this field of engineering and science and cutting-edge science, AI, the kind of cutting-edge science that raises, for some people, maybe existential questions. about life and what is our role and maybe even like the coming of transhumanism and um you know terminator like oh the the missiles are going to be launched tomorrow

8:45 and so maybe i better get right with jesus because the end might be coming and the computers will do it So as a Christian person, an intelligent, thoughtful, knowledgeable Christian person, I would be interested to hear how – what parts of what you’re doing do you say are being informed by your Christian faith?

9:12 Or is it just I got my Christian faith over here, and I got my AI algorithms over here, and never the two shall meet?

9:19 Well, you know, I’d like to hear the opinion of an intelligent, thoughtful Christian person in the space as well. But since I’m here, I’ll do my best. Okay. That’s all we can ask. Yeah. So. All right. So. God created us in his image. He didn’t create computers in his image.

9:44 So one of the problems that people get into when they’re early on in developing AI or using AI is they anthropomorphize it. So saying artificial intelligence is shorthand. for saying what it really is because saying massively piecewise linear regression or optimization functions that use lots of little nodes to come up with a hybrid or ensemble to answer,

10:23 nobody really feels that, right? That’s a difficult thing to feel and immediately kind of get what that means. But if I say artificial intelligence, then suddenly like, ah, it’s it’s Wally. It’s it’s Hal. I kind of understand where we’re going with this if we say that it’s artificial intelligence. And so one of the hazards, though,

10:52 is there’s a difference between that science fiction, which really truly anthropomorphizes it, gives the AI motivation, and the technology that we actually have. AI doesn’t have any internal motivation other than to solve for the metric that you gave it. Whereas God created us in his image with a purpose to glorify him.

11:23 And so there’s all sorts of differences there just on the base level of what’s going on mechanically in the background.

11:35 OK, I’m still with you.

11:37 Yeah. So when it comes to the day to day of working on A.I. as a Christian person, I can view it as another way to program computers. It’s just simply another way to put something in and get something out and make a computer program that’s gonna be useful for humanity.

12:07 If I start viewing it as an entity, then I start building technical debt into whatever system I’ve built out to where I’m okay with not understanding how the back end of it really works. And then I get into some tricky situations where I get unexpected results,

12:29 but for some reason I believe it or I trust it and think it has some secret knowledge that I don’t have. when there is no secret knowledge there, it’s still a computer program, it’s still garbage in, garbage out. So as a Christian AI scientist, my viewpoint is the AI is still a computer program, it’s still garbage in,

12:49 garbage out. Whereas if I didn’t have that viewpoint and I attributed to AI the same sorts of motivations and value that I attribute to humans, then I can start to misunderstand what I’m getting with with the AI. And we hear of these situations where an AI engineer thinks that he’s created a real intelligence.

13:17 Right, right.

13:17 Yes. But as a Christian person, I look at that and go, you wrote a computer program. You told it to tell you that it has feelings. It told you that it has feelings. And because you don’t have that external frame of reference that God created man to be in his image,

13:37 God didn’t create the computer to be in his image, you lost your way and you started believing that. You started believing the computer program that you told to tell you that it has feelings. And now you start to believe that it has feelings. It’s like, there’s a reason why chatbots work better when you’re polite to them.

14:00 It’s because they’ve been trained on a lot of data where people that were polite got more out of the system than people that were rude. And if you trained

14:11 That’s a great pro tip for me sometimes.

14:14 I mean, if you trained the chatbots with data of, you know, data just from like Dave and Busters or things where people that were rude got more out of the system, then you would have to be more rude to the chatbots to get something up. But it’s still garbage in, garbage out.

14:33 And whatever you’re putting in, it’s just a mirror. And that’s what you’re going to get out of it.

14:38 Okay, okay, so we’re cruising along here. So one of the things, I had this conversation a couple times this past week in Washington, I was talking to Two different folks back to back, both were having, well, one of them asked me, this person said, you know,

14:57 I’m considering leaving the government sector and going to work in the private sector. And in the private sector, I would potentially be working for a defense contractor, which is not at all unique here in Washington that people leave to go to work for defense contractors. And then they said, well, you know, here’s my question.

15:16 The thing that I’m having to wrestle with is this particular defense contractor does manufacturers things, devices with guidance systems that are designed to go somewhere and blow up and kill people. And I’m trying to figure out how do I reconcile that with my Christian faith? Because part of my Christian faith would tell me like, hey, you know,

15:48 it sounds like what you’re trying to do is be the best person you can possibly be at killing people. And I said, well, you know, actually, you know, you’re not far away from the truth on that. The key here, though, is understanding that God cares very much about the way in which we engage in warfare.

16:11 It’s not that God wants there to be warfare, but when we engage in it, he cares about what’s happening there. And so if you’re going to leave the private sector or the government sector and go to the private sector, then God is gonna care very much that you see your programming and development of

16:29 those guidance systems as an act of worship. And therefore you attend to it with great care so that those guidance systems take it exactly where it’s supposed to go and detonate with only the yield that it’s supposed to detonate with. And that only the right people are being given information how to use that guidance system.

16:52 Because those things matter. And God cares very much about guidance systems not being sloppy, sold to sloppy people, used in sloppy. That’s not the way that God wants to be worshipped.

17:06 That makes a lot of sense. Go ahead. What I’m going to do is just hop in there and just kind of say that we take a very similar approach, not just me, but the whole company takes a very similar approach and has a missional focus. Uh, certainly not everybody in the company is a Christian.

17:32 There’s some other, other folks in the company that are, are Christians that I work with. Um, But the overall mission of our company is to make the world a safer place. And so we very much limit ourselves to the technology that we see can successfully make our warfighter safer and not unintentionally cause collateral damage and

18:02 things of that nature. So your description is in line with how we view it. If we feel like a technology that we’re building has too much collateral damage, our collateral damage tolerance is zero, then that’s not something that we’re gonna engage in at this time. if it’s a different situation where it’s like a battlefield battlefield situation

18:27 everybody there has agreed that they’re going to be in danger already that’s a little bit different conversation and you know we can have that uh that conversation but But ultimately, that’s our focus, and that’s how we align those two things, if that makes sense.

18:46 Yes. So in Faithful Presidents, we talk about these four different questions. What is good that I can encourage? What is broken that I could have a role in restoring? What is missing that could be created that would be a benefit? And what is evil that I should oppose?

19:06 And as we begin to think about AI, I think all four of those questions are very relevant. or how we approach AI. So when we talk about things, let’s just start at the top, what is good that I can help to improve or encourage? I go through airport security at least once a month, right?

19:30 And so I’m TSA pre and all that stuff, but it has gotten better. It has gotten, I just saw a thing on our local news that this year, In Washington, D.C., in all of 2024, they confiscated over 400 handguns trying to go through security. And one of the things I found interesting is they said, oh, well,

19:58 the number one response of the person who they – hey, you have a gun in your carry-on bag. The number one response they have is, oh – i totally forgot i had it in there which to me is like yeah you know what you don’t

20:13 ever get a gun again if you don’t know where your gun is at any given time you probably shouldn’t have one um but so like i know one of the things that analytical ai does is you’re working with uh tsa if i if i write remember this right to help help

20:32 Those decision makers who are looking at the screen saying, do I stop this person? Do I let this go? What is this? What’s going on here? Talk a little bit about how your company helps TSA, which is a national security concern. How do you help them do their job?

20:49 Yeah. So, um, there’s a lot of details that I can’t go into, of course, of course, but, uh, but what I can say is we’re working with TSA on the next generation on-person screening and baggage screening technologies, uh, to have algorithms which do high detection with low false alarm, um,

21:13 both for on-person things as well as for luggage, um, and ensuring like our on-person things are gender agnostic so that somebody walking through doesn’t have to, to click whether or not they think that the person is male or female and then have the chance of their personal identification with those things be different. Right. Um,

21:38 as well as a variety of other applications. And that’s a great example of a place where traditional mathematical computer algorithms would never be able to get us where we need to go with the detection. So the detection required is above and beyond what traditional math and signal processing could get us.

22:10 But using AI algorithms where we’re essentially showing it examples and then saying, this is what bad stuff looks like. This is what good stuff looks like. That’s something that AI can do and it can do well. And so that’s the space that we’re in with TSA and honestly with Customs and Border Protection as well,

22:36 because preventing narcotics coming through the border is huge. And so that’s an area that we’re working with as well to ensure that there’s a material dent in the amount of fentanyl that’s coming into the country through ports of entry.

22:53 you

22:53 Yeah, I’m all for that. So here’s one of the things that I wanted to just point out in what you said is one of the things that is good that needs encouragement is we need to help the people who are trying to stop the fentanyl from coming in.

23:13 We want to get as much of that fentanyl out as possible. We want to stop people who are trying to take weapons and whatever in their luggage. There are people who are trying to do that as a good, and we want to help them. The other thing that you said,

23:26 which is to lower the amount of false flags, is the thing that I’m always encouraged about, or should I say, the thing that makes me discouraged is when I go through a security checkpoint, and I usually have basically the same thing I take with me almost all the time,

23:47 and occasionally somebody will stop me and they’ll flag my bag and it goes on the second track. I’ll say, can you tell me what flagged it? Oh, yeah, this thing. And it’s something that I will always bring through. I always bring it through. But the person says, oh, well, he’s new.

24:07 And so he’s still learning whether or not he should stop this thing. But of course, I open it up. I look and, oh, it’s a little cylinder disk. And it’s your shaving soap. So he’s like, oh, yeah, OK, I know what that’s shaving soap. Yeah. But that’s inconvenient for me.

24:23 And because I’m a upper middle class white person living in the West, being inconvenienced for 30 seconds going through TSA, it’s the kind of thing I don’t ever wanna have to experience. But it is a good thing when a person’s traveling on Christmas or Thanksgiving or

24:44 something and they’re trying to make a plane and they’re not being stopped. when they shouldn’t be stopped because they’re like, hey, we can get more people through here faster when the artificial intelligence is scanning things faster than he can scan them. and is going to stop the person they need to, but not stop the other people.

25:07 So there’s actually two things happening here that’s good. One is stopping the evil and the other is letting people with their families go through.

25:18 99% of people, there’s no reason to stop them. And so if we can, with our AI algorithms, process that 99% and say, nothing to see here, then that’s great. And that is on the TSA roadmap that’s published to get to image on alarm only, which would be exactly that,

25:41 to just let just the AI algorithm look at everything and then only flag the things that look potentially problematic for human review. We’re a bit off from that, but But it’s on the TSA roadmap to get there, and we’re working with them to move it that direction.

26:00 Yeah. And I think you said something really important there is, one, 99% or more of people who are flying should not be stopped. They’re doing fine things. And they do catch most of the stuff. They do catch most. But there are still a lot of things that get through there. And thankfully,

26:21 because 99% of the people aren’t intending to do anything bad, the thing that they’re bringing through that they probably shouldn’t have wouldn’t be an issue anyway. but AI is something that is going to bless those TSA agents. Like they are going to be very, there’s no TSA agents going, boy,

26:41 I sure hope AI doesn’t ever take my job.

26:44 And I think the space that I like to play in is human in the loop applications. So AI, as I said earlier, AI doesn’t have a motivation. It doesn’t have context. And so it’s generally a mistake in high trust situations to take the training wheels off and let AI make the decisions.

27:11 But if you can have it be essentially a really big lever or a really big hammer so that the person in the field can get more done, then that’s where it really can be a blessing and a force for good. So when our algorithms make it to the field,

27:34 not a single TSO is going to be taken off the line. We’re having a hard time finding, you know, as with most jobs these days, it’s difficult to find enough people with the right backgrounds to fulfill those roles. And so if we can make the people that are there more effective, we’re not taking anybody’s job.

27:58 We’re just making them better at their job.

28:00 Okay.

28:01 There was a study. I don’t know exactly what it was, but even I think it was done by the Boston Consulting Group. And they used like 700 of their consultants because they’re massive. And they kind of split them into two groups. One group that was using the AI and another group that wasn’t using the AI to solve

28:25 their problems. And what they found was… the group that wasn’t using their AI had a pretty broad distribution of effectiveness. You had your top performers, you had your low performers. What they found with the group that was using AI was that everybody got better, but the people that maybe wouldn’t normally do as well,

28:50 they would be closer to the bottom of the spectrum, they got pushed up more than the top performers got pushed up. Oh, that’s great. That’s fantastic. Essentially, the AI just it was kind of it leveled the playing field quite a bit so that people that understood how to use the AI were operating in top performer space where

29:15 previously they they weren’t there. And so that’s what I’m talking about, of using that AI as a big lever, a big force multiplier to get more things done that you wouldn’t ordinarily be able to get done by yourself.

29:33 Okay, so I wanna talk about the next two questions. I’m gonna set up the question by saying, I’m gonna introduce what I would consider to be a category error. And we’re gonna use this example that I know has been in the news. So the question is,

29:52 how do we think about AI in terms of what is broken that needs to be repaired and what is evil that needs to be opposed? So there’s a couple news stories floating around out there about the latest generation of chat GPT and it being used for basically online therapy. Like just talk to the chat bot, right?

30:17 Tell your problems to chat bot, chat bot will respond back. And, you know, they loaded up whatever information, but the stories that make the news are the ones where the AI chatbot suggests that the best solution to this problem is either you killing yourself or you killing your parents, right? Do you just kill them?

30:39 And of course, so the category error, I would say, is saying that that’s evil, that that the, and this goes back to your anthropomorphization comment, that that is to attribute to AI something that it does not have the capacity to do, which is try to do evil, try, because that’s a human thing. We do the evil.

31:07 And so it’s not that’s not the evil thing. It’s a broken thing. It’s that the AI is broken. And maybe you could talk a little bit because I know this is something that you guys you guys think about is is, oh, AI only knows what you tell it. And you’ve used this phrase multiple times.

31:24 You take the guardrails off and we have to be careful about that. So talk about how something like that happens. and why AI data scientists, how they rectify that maybe?

31:39 Yeah, so for context, there’s lots of things that we interact with that we feel like our AI, which are really what we call ensembles or groups of things, okay? So like if you have an Alexa device in your home or a Google device in your home and you just say, hey, Google or Alexa,

32:04 there is one AI model that is listening for you to say Alexa or hey, Google. There is another AI model that’s going to transcribe what you said into a sentence and then pipe that into something else to figure out what question. There’s another model that takes those questions and matches it with the most appropriate domain.

32:30 And then there’s going to be another model that’s going to go out and find whatever the answer is and then serve that to you. So these things get ensembled pretty broadly. now what’s happening in the case of those tragedies because i think that’s the only

32:46 way to to think about them those tragedies where somebody gave the ai the power over their emotions and their universe to and believed it is that somebody built an ai model based on therapy transcripts they They probably included both the patient and the therapist’s pieces. Probably didn’t give the AI model adequate differentiation as to what responses

33:26 were appropriate. and then probably didn’t put an additional check in there to see whether or not there was harmful, something harmful coming out of it. So if you go to the production version of ChatGPT on OpenAI’s website or Copilot, if I ask it to give me a picture of two robots sword fighting, It may not do that.

33:57 It may say this violates our content standards. We can’t do that. You’re going to need to rephrase your question. And that’s because they have the underlying model. And it’s interesting because you can ask the question. It’ll say, let me try to do that for you. And then it’ll come back and say, oh, wait, I can’t do that.

34:15 And so that’s exposing that what’s going on is there’s a model that’s taking your request. There’s another model that’s trying to create the picture that you asked for. And then there’s another model on the back end that’s in parallel checking the things that you’re asking for and saying, is this harmful? Yes, no.

34:35 And so what’s going on in those cases with therapy is somebody was excited by the technology. They trained a model that showed initial promise and They did not conceive of the series of events that would lead to the request for this person to kill themselves. And so they didn’t build those guardrails in place.

35:03 And so this is where that technical debt that I referred to comes into place earlier. And what I characterize these sorts of issues is broad brushstrokes, a thing that all AI ends up having a problem with at some point is bias.

35:24 Okay, explain that.

35:25 So bias is simply your training data led your model to make some assumptions about the questions that you’re asking it that you didn’t intend just because of the way that you set up your situation. So this has happened historically with scenarios where people are building financial models to determine who should get loans and who shouldn’t.

35:55 The AI model may just learn that people of various ethnicities shouldn’t be granted loans, not realizing that really the thing that it should have learned was people with certain patterns of credit history should be denied loans.

36:12 Okay, so you’re talking here about correlation and causality, right?

36:16 Yes, and AI has no notion of causality. It only knows correlation. And I’m happy to argue with anybody that would say that AI can learn causation because humans have a problem with that. You know, I love the websites of the spurious correlations where it’s like how much butter is consumed.

36:39 versus how many Nicolas Cage movies there were that year. Those are great, but if you gave AI that data set, all it would tell you is, oh, a lot of butter was consumed, we better get ready for a lot of Nicolas Cage movies. And so what happens is things like facial recognition, right? Facial recognition,

37:03 most of those models were designed and the data sets were curated by white nerds in California. So surprise, surprise, the people that it’s good at recognizing look like white nerds in California. And if you don’t match that demographic, suddenly you’re a single mother in New York of color and you’ve been put in jail

37:25 overnight because the police didn’t believe your driver’s license. They believed the AI model that was designed that was inherently broken because of this bias due to improper training and vetting of, of the models. Um, So those are some examples, but that’s essentially what happens with these models. As they get more complex,

37:54 our ability as humans to know the ways in which that bias is going to play out becomes a lot more difficult to track. And so that’s where it gets broken as we build up this technical debt. There’s a paper that a scientist at Google wrote, I think in 2015,

38:17 that called machine learning the high interest credit card of technical debt. Because you can just put data in, you get something out, it looks really promising, and then you can be off to the races and have sold it to a bunch of people before you even realize where you made your mistakes.

38:37 Okay, so let’s keep going down this road because this is – and this is the thing that I think I would say maybe scares folks because we’ve all watched the movies with Will Smith where he’s got to try to find the one computer that’s like trying to turn all the other computers evil.

38:57 They’re just coming out like crazy now, right? Yeah. But we also know that there are state actors who embed information and malware into water treatment facilities, telecommunications facilities. All of this stuff is going on, right? AI is trying to generate passwords faster than we can so that it can hack things. So…

39:25 When we go back to this categorization of what is broken, what is missing, and what is evil, this is really where the evil of AI is really in the hands of the beholder, meaning the person who controls it is going to make the decisions about what to do with it, which then begs this question.

39:48 How do we as Christians who have some moral basis for the ways that we make decisions, how do we develop AI knowing that there are people who are not virtuous, who maybe wish to do us harm, or maybe they’re just complete anarchists and they don’t care about anything.

40:12 They just want to, as the Joker said, some people just want to watch the world burn, right? How do we think about AI in those settings? Because I think that’s the thing that concerns people is people intentionally applying ai to accomplish really wicked things yeah and and i

40:37 often say you know we don’t get to mad max fury road because of a nuclear war we get to it because social media and because somebody read something that an answer to a question that they gave AI, and then they believed it. Because it’s kind of a mirror. So if you have a psyche like that,

41:08 that just wants to watch the world burn, you can get answers that tell you, yes, the world should burn, and then develop an almost religious-like awe for the AI and for the information it’s putting out. So I think at the end of the day, if I had to really categorize it in one sentence is,

41:38 AI can’t be our idol. And any place we see AI becoming someone’s idol, it’s where they go to for all of their information. It’s where they put all of their time and energy, all this kind of stuff, all the normal things that you use to tick off, whether or not something’s an idol in your life.

42:02 That’s where we can start to be worried. I think it’s a long time before AI is going to have enough access and enough complexity to start hacking all of the computers and launching nuclear weapons and all that kind of stuff. I think it’s much more likely that a powerful CEO that

42:35 owns an electric car company or something of that nature, asks the AI a question, decides that all the humans are evil, and he goes out and changes a configuration file. And then suddenly all of the electric automated driving cars are running into people instead of avoiding people.

42:58 Ah, so now you’re putting you, meaning data scientists who have access to these kinds of things, you’re saying we matter, meaning how we think about the world and how we appropriate ourselves and how we think about humanity matters very much. for a data scientist because they could cause existential crisis in this world.

43:29 Yeah, absolutely. I’ve, I’ve heard of, I have friends or family that I’ve heard things of, of, yeah, you know, this, I, as my career was growing, I wrote this program and, um, And everybody was really obnoxious to me. And so I wrote something into it so that once a year it would break and they would

43:54 need to call me and I would come and I would just enter something and then it would work again. Nice.

44:00 That’s good.

44:01 And, and you can imagine somebody that doesn’t have the viewpoint of people are born in the image of God. overly anthropomorphizing the the ai giving it more credence than it needs being that engineer that designed something to tell me it has feelings and then it says

44:24 it has feelings and i believe it and and then it tells me to do something and then I do it, all of those sorts of scenarios become a real problem. And right now that switch, the person that has the real power there is the person that’s interacting with the AI model, not the AI itself.

44:48 Okay. All right. So because Faithful Presence cares about the public square, we try to relate things to the public square. So that’s what we’re going to try to do now. So I think it’s on Monday is maybe it’s Sunday. The clock runs out for TikTok. They’re supposed to, you know, divest or be shut down.

45:06 But then as a friend of mine is an attorney and he is one of, I don’t know, tens of thousands of attorneys that TikTok has hired in the United States to try to like help us out. um and then i just saw this past week that the biden administration says well you

45:25 know maybe we’re not going to be so eager to kind of you know get rid of maybe figure out some way to make this thing happen and then i saw that the ceo of tiktok is going to is going to be at the inauguration so there’s you know it’s all these

45:37 conspiracy theories all this other stuff out there but what i want to do is is not necessarily talk about the conspiracy theories but just to talk about the reality is that People went to their congressmen, and Jonathan Haidt has written a wonderful book called The Anxious Generation,

45:55 where he talks about the impact of social media and computer screens and things like that on youth, which we’ve talked about before on the podcast. But eventually, this issue got to Congress about, hey, TikTok bad. But it really has nothing to do with the social media aspect of it. We’re concerned that China is getting our information.

46:20 That’s the real issue here. The social media thing, whether or not you can dance and your mom can dance and sell chocolate bars, we don’t care about that. We just want the Chinese to have your information. But – What I found interesting was in these conversations when Mark Zuckerberg and some

46:39 of these other people have been before Congress, watching 70- and 80-year-old congressmen and senators try to ask questions about a technology they clearly have never used and don’t understand at all, they’re like completely guessing. How are we supposed to look into the public square and say, okay, I think we need some guardrails. I think we,

47:08 I don’t know that we can trust ChatGBT and Facebook and Meta to just do all this thing and be virtuous. Our senators and congressmen are gonna have to do something. And they don’t, I mean, I don’t know anything. You don’t want me making these decisions.

47:26 How do we think about AI as a policy that needs public square consideration? Does that question make sense?

47:38 Yeah, and I think, again, it’s good to remember that within our Christian framework, on his own, Man is evil. And so while capitalism is great, it can’t solve the problems for us without us going back to remembering that without Jesus we’re evil. Our intent is not to do evil. Good. That’s why the EPA exists,

48:26 because if you have a situation where your negative externalities to your process can be pushed onto somebody else, then capitalism says, sure, go ahead and put that hidden cost on somebody else and you profit from it. And so what happens in the case of AI is those negative externalities have yet to

48:50 be realized or they’re being realized and we haven’t as humans recognize what those negative externalities are. I mean, I think that a lot of popular AI models are examples of the greatest heists of our modern era, because the underlying data to develop those was essentially just stolen off the internet because people left their Facebook profiles public.

49:20 Huh. This is called scraping, right? This is what scraping is?

49:24 Yeah, yeah. Web scraping, right? I can set up a Python computer program today that’ll just scrape the internet for all sorts of information that I don’t own. And the overarching logic is, well, it was publicly available. And it’s like, well, I can go to the library and photocop Dickens, but that doesn’t mean I own it.

49:50 Okay, good point, good point. And so all of these things, and then even the social media companies themselves make it very difficult for you to understand what they can and can’t do with your information. And then they also change it all the time. without telling you and without providing notifications of what those things are. So broad brushstrokes,

50:26 I think it would be difficult to say that the free market is going to regulate this properly because social media companies have already decided that what they want to do is keep people on their platform for as long as they can, regardless of what that means for the content of the platform. And so

50:55 things like Facebook become everybody’s own individual Fox News or CNN that just keeps them engaged, feeds them just enough stuff that they really disagree with so that they wanna get on and debate it. And then feeds them a lot of stuff that they agree with And then the moment that you’re starting to get disinterested, then, you know,

51:18 one of your friends from an opposing political viewpoint suddenly has a conversation they’re having on Facebook, and then you’re reengaged. And they have no incentive to… do anything different than that, because that’s what’s keeping people on the platform. That’s what’s providing them with their ad revenue streams and all that.

51:44 And a lot of that ad revenue streams is all developed on the back of selling your demographic information. to other people. That’s the free market exchange that I’m making to be on Facebook is I’m selling the fact that I’m interested in guitars and that I’m a male in my 30s and all of

52:05 these things in exchange to be able to share my kids’ photos with friends and family and grandparents and such. And so I think it’s challenging to say that AI developers are going to put these guardrails on themselves. I think in a lot of cases, there’s good intentions too.

52:31 I think OpenAI as a company has at least attempted to say that they’re putting those guardrails on. Even like Google with their models has. But then you look at the history of those companies. Google’s motto used to be don’t be evil.

52:53 And then I don’t know where that motto went, but I haven’t heard that in a long time. Don’t be evil. I like that. That’s good. Open AI used to be open. And then the moment that they created the biggest AI breakthrough in the last 10 years, suddenly they’re not open anymore. They’ve just retained the name.

53:19 Yeah, they’ve retained the name because that’s their branding. But… But yeah, they went from trying to make sure that Skynet never happened, and then the moment that they kind of got close to something that kind of feels Skynet-y, The wall came down.

53:39 Right. Which just only adds to the fear, right? That adds to people saying, oh, see, something bad is going on.

53:47 And I don’t know that it’s something bad. I think they just realize, oh, we have something that’s potentially monetizable. Right. And humans are driven by those things. incentives. And so the incentives are to figure out how to monetize that and not deal with the negative externalities of having scraped the internet for all that information and

54:11 push those negative externalities on somebody else.

54:14 So how do we How do you, when you look at the people that are currently in Congress and you say, okay, and I’ve got to imagine that analytical AI either has their own or pays money to somebody to come in and lobby or advocate for AI. Tell me what AI advocacy is.

54:41 looks like like what kinds of things are you like okay we like we need to make sure that congressmen and senators know this like this is what is good this is what is broken this is what is missing this is what is evil i’m imagining that’s what your

54:53 advocacy people are doing right yes so so we focus on a few different things when we’re talking about um what, what should be looked at. Um, one is certainly AI explainability. You know, if, if you look that up, I don’t know if it’s actual word,

55:13 but basically can you query the AI for why did you make that determination and get some kind of sensical response? Um, and transparency as part of that as well. The other thing is for ethical data collection, because, because it’s very, and a lot of government spaces have already started pushing on especially the

55:42 ethical data collection piece because they’re realizing like that is problematic. If you’re just scraping the web for pictures of people and then training your AI on it, history’s not gonna, that’s not gonna be the right side of history. You’re gonna want to have permission from the individuals that you have those images of that says, yes,

56:11 I’m okay with that being used for AI training or whatever it is. So I think those are some big steps. And then also just the applications themselves. And like, Is this a high trust enough situation that a human should be involved in this decision? What’s the worst thing that could happen if I get a false alarm, right?

56:43 So if I’m building a medical AI model, let’s say, and it’s diagnosing cancer, a false diagnosis of cancer in some cases can be just as harmful as getting a true diagnosis of cancer. It’s a life-changing event regardless of whether or not that diagnosis is true.

57:11 And so we also have to be careful with the applications that we’re using. We like to draw the analogy. So we do those algorithms for TSA. We also have some projects with the Army. If I tell a TSA agent that you have a gun on you, what’s the worst thing that’s gonna happen?

57:37 You’re gonna get groped by a stranger. If I tell somebody in an armored personnel carrier from the army that you have a gun on you and you look like a bad guy, you’re not gonna be here anymore. And it doesn’t matter that you’re carrying a broom.

57:56 It doesn’t matter that you may be a kid playing with a stick The problem or the negative impact of a false alarm just skyrockets depending on the application. And so I think those are all conversations that we need to be having is what’s the benefit if it gets it right?

58:22 What’s the negative impact if it gets it wrong? And how can I as the human… ensure that even if it gets it wrong, I have enough information that somebody can bear responsibility for what happened. Because the AI can’t bear responsibility. It has no motivation. Like you said,

58:46 we can’t even ascribe evil to an AI system that tells somebody to kill themselves. We need to have enough information there that the human that’s there managing that AI system can bear responsibility for whatever the outcome that the AI system is providing.

59:11 Yeah. Okay. So this has been like, I get to do these and they’re always fascinating, but this one, I feel like I could go on for like three hours. This has been great. And I appreciate it. The thing that I appreciate the most about this conversation is helping me, helping us, helping our listeners,

59:33 who are both public servants here in Washington, as well as pastors and people attending churches, who are like, ah, just tell me about the public square things. But one of the things, I was talking to somebody just earlier today, and they’re like, okay, what do you do? Like, what do you do? I’m like, well,

59:51 we do a lot of things, but one of the things that we try to do is help people talk intelligently about public policy issues and not to make straw men out of them and either attack them or run away from them, but actually understand that. And I just really want to say thank you so much,

1:00:10 first of all, for what you do as a data scientist. and helping me get through TSA quickly. I very much appreciate that. But also making sure the wrong people don’t get through. But also, thank you for helping us understand AI, why it matters, how there are things about it that are good, things about it that are broken,

1:00:33 things about it that are missing, and things about it that can be used for evil. And I just I really appreciate just the breadth of your knowledge and your ability to help us think about how to apply our Christian faith to the issue of AI and flourishing. So thank you for that, my friend.

1:00:55 Yeah, it’s been an absolute pleasure, Michael. And if anybody reaches out to you and wants to talk more about it, I can’t promise that I’ll respond to an email, but you can certainly connect us up.

1:01:09 We usually try to end the podcast. I’ll ask, is there anything we could be praying for you for?

1:01:15 Um, yeah, it’s just it’s it’s busy. It’s difficult to there’s just so much context switching. We’re we’re growing as an innovative company in this space, as you can imagine. And so it’s it’s difficult to keep up with everything personally. the just being a father and trying to parent wisely and well in this ever-changing

1:01:45 space that we find ourselves in right now is is a challenge um so just for uh energy and emotional support for uh for me and alicia would would be great um and certainly wisdom uh wisdom and parenting all of those things are are

1:02:05 definitely things that we need okay but if you think of other things you want to

1:02:09 pray for i’m not going to stop you okay um how’s the music i know i know it’s a side gig but you are one of the most talented musicians i’ve i’ve ever known personally um how’s that going you still you still keeping up with the music thing

1:02:25 I still play a good bit of music. It’s been a while since I’ve gigged necessarily. But Alicia and I lead the music at our local church that we attend, Cross Creek Church, down here in Birmingham, Alabama. And so that’s fun. I still get to pull out the dobro. And, and, and that sort of thing.

1:02:51 I think on, on Christmas Eve, I was playing the banjo. So that was fun.

1:02:56 Okay, good.

1:02:57 So, so yeah, certainly still doing a good bit of that.

1:03:01 OK, well, well, thank you very much, Will. I appreciate you joining us today and you have every every blessing that I can send your way.

1:03:12 Thank you.

1:03:13 Thank you very much for joining us. Thanks for tuning in. We’re really thankful to have had Will Monroe on with us today. That’s it for this episode of the Faithful Presence podcast. Do pray for Will, pray for analytical AI, pray for other data scientists, pray for work that’s happening in the AI industry,

1:03:36 and pray for those folks who are in the public square who are trying to figure out what does appropriate guardrails look like and how do we continue the flourishing process of our society and incorporate AI into that, but do it in a free and virtuous way. If you enjoyed this podcast,

1:03:59 we encourage you to give us a like, give us a share, give us a rating, give us a review. It helps all the algorithm thingies that we just talked about work. Also, thanks to our team. Thanks to our chief of staff, Josie. Thanks to Andrew, our new podcasting engineer for making

1:04:18 These podcasts at least appear visually and audibly appealing. And now receive this blessing. May our father in heaven make you one as he and his son are one so that the world will know that he loves them and sent his son for them. Go forth in peace in a posture of faithful presence. Bye now.

You can also view the video version on our YouTube channel.

Work Cited

  • NPR story on Character.AI recommending a child murder their parents over screen time restrictions. https://www.npr.org/2024/12/10/nx-s1-5222574/kids-character-ai-lawsuit

  • American Counseling Association, Recommendations For Client Use And Caution Of Artificial Intelligence. https://www.counseling.org/resources/research-reports/artificial-intelligence-counseling/recommendations-for-client-use-and-caution-of-artificial-intelligence

  • Machine Learning: The High-Interest Credit Card of Technical Debts, https://static.googleusercontent.com/media/research.google.com/en//pubs/archive/43146.pdf

  • New York Times, Clearview AI Used Your Face. Now, You May Get a Stake in the Company. https://www.nytimes.com/2024/06/13/business/clearview-ai-facial-recognition-settlement.html

    ——————————————————

“B-O-O-G-A-L-O-O” (2010) by the Diplomats of Solid Sound is used by license. Scheduling by Josie Risinger
Post-production by Andrew Brinkmann
(c) 2025, Faithful Presence, all rights reserved

Faithful Presence exists to provide whole-life discipleship in the whole of life for the whole of the public square. Join us as we seek renewal in politics and the public square by becoming a Supporting Partner.

Find Us On: LinkedIn | Instagram | Facebook | X | YouTube | Apple & Spotify

Share this article by email