Algorithms vs Souls: The Use of AI in Global Missions

[00:00:00] Speaker A: Welcome to ID the Future. I'm Andrew McDermott. [00:00:03] Speaker B: Today's episode comes to us from our [00:00:05] Speaker A: sister podcast, Mind Matters News, a production of the Discovery Institute's Walter Bradley center for Natural and Artificial Intelligence. [00:00:14] Speaker B: You can learn more about the show and access other episodes at mindmatters. [00:00:19] Speaker A: AI. [00:00:24] Speaker C: Greetings and welcome to Mind Matters News. I'm your host with Real Intelligence, Robert J. Marks. You know, AI is having an impact everywhere, but what about Christian missions? How are those in the mission field using artificial intelligence? Our guest today is Dr. Don Barger. He is the Director of Innovation and Artificial Intelligence at the International Mission Board. Don works with new technology, especially AI, and he helps people use it in smart and responsible ways. We're going to be talking about that today. He thinks a lot about AI can help in real life situations, especially in global missions. He also understands the limits of AI and what it can and cannot do. We're glad to have him here to talk about how AI is changing the world and what it means for the future. Helping me out today is Jonathan Swindell. Jonathan is a contributor to Mind Matters News. He's a computer engineer and is a PhD candidate at Baylor University. So he'll be acting as my co host today. So let's start out with Don's PhD. I'm going to give it over to Jonathan now because John's PhD is in an area I never heard of before. [00:01:32] Speaker A: Yeah. So, Don, I believe your PhD is in orality, right? [00:01:38] Speaker B: It is. And a lot of people think morality, but it's not morality, it's orality. [00:01:43] Speaker A: So tell us, what does orality mean? [00:01:45] Speaker B: You know, it's just recognizing that people get information in a variety of different ways. And if you go back in history, well, first of all, I just had a grandson born on last Friday. Henry is new into the world. And you know what Henry is? Henry is an oral learner. Right now, Henry is an oral learner. He will be an oral learner for the next few years. Everything he learns will be through observation and what he pulls into his senses. He won't be a reader. He's going to be really smart, no doubt, but he will not be reading in his first few years of life. And we, all, those of us who have learned to be literate, have acquired that ability to take in information from a non oral modality. But the reality is all of us at one point were oral learners. And my research was in how we could use the same principles we use for Bible translation. The clear, accurate, natural. How can you take that? And to make sure that Bible stories, as we're talking and we're sharing with oral cultures, how can you do that in such a way that you make sure that the story is clear and accurate and natural? [00:02:57] Speaker A: So can you help us understand the difference between an oral culture and a culture with a high emphasis on literacy? I don't know if that. That's not something that we typically think as much about in an AI super tech bubble. So we have a lot to learn from people, you know, working in this sort of linguistics, cultural setting in how we can apply these things. [00:03:20] Speaker B: We're almost in a. It's interesting. I wrote a book with Dr. Grant Lovejoy a couple years ago called Unreadable, another book you probably won't read. And the principle of the book is most people get information from non print resources. And I think this little device right here probably has more influence on us than anything and the way we read. And I know this isn't exactly oral, but there's a term that's coined called digitoral. So it's kind of combining the digital and the oral. So even for those of us in learning environments that are very first world in academic institutions, the number of people who get information from this very medium that we're in right now, this is a podcast. This is how people are getting information. They're exchanging ideas, this idea of information flow. And so in the research that we did, we interviewed several hundred people all over the world about how do you get information that shapes who you are and what you believe? And it's amazing. The number of people who told us they got information from books was exactly one out of all the research that we did. So that doesn't mean that people don't read. We actually wrote the book, which is ironic that you write a book about not reading, but the book being published gave us legitimacy of the research that we did. And so I'm not saying that people shouldn't read. I'm all for people reading. But we have to recognize that there's a lot of other medium out there that helps people get information. Storytelling is a traditional way that a lot of cultures share information, get information from one person to the other, handed down over generations. [00:05:11] Speaker C: Oh, Don, let me ask you this. I think this is a criticism that some other religions, like the Muslims have of Christianity. They say that passing down things with oral tradition, whereas you tell your kids, they tell their kids, they tell their kids, is something like playing telephone when you were a kid where you lean over and you whisper to the neighbor next to you, some sort of phrase and he whispers it to somebody else. And then when you get to the end of the chain, it's totally, totally different, right? Yeah, it's totally different. So clearly there's a difference here. Could you unpack that a little? [00:05:42] Speaker B: Yeah. So I've heard this criticism a lot. And what's interesting is if you think about the game telegram and most are telephone. Most of us have played this game. I started in youth group. You would lean over and you would whisper ever so carefully and you were only allowed to tell it one time. You would tell one person and they had to get what they got that one time. And it was usually some weird, esoteric, weird story that was really hard to remember. And then they had to immediately tell it to somebody else who told someone else. Well, yeah, that's really hard to do. And I would say that the game itself is set up to help you fail. Oral cultures, on the other hand, are very different than that. Usually it's a community knowledge. And so it's not like some secret knowledge that only I have. And I'm going to whisper it as quietly as I can and as quickly as I can. And you've got to immediately tell it. Actually in oral cultures, like the story itself is told by the community over and over and over again. And the moment that someone introduces something different to the story, the whole community recognizes. Whoa, whoa, whoa, that's not the story. I've got a real example of this. My family and I, we lived overseas for 24 years and we were missionaries and we served of in some small people groups that were primarily oral communicators. They could read some, but not very well and they didn't prefer to read most people, but we would tell Bible stories. And I remember one time there was a guy, we will call him Jay, just to protect him because he was notoriously bad at telling stories. And so we would tell a story and this was a Bible story. And I remember that he put like he would get names mixed up and once he talked about Jonah on the ark and immediately, immediately when he said that, the entire group, even the little kids spoke up and said, no, no, no, it wasn't him, it was Noah on the ark. And so there's this self correction that happens in these oral communities where the story itself gets preserved by the whole community. So exactly the opposite of what you have in the telephone game where it's kind of set up to fail. This is actually giving you every protection you can to make sure that the story is preserved and doesn't change in [00:08:06] Speaker C: talking to you about this. Prior to our conversation here, it occurred to me this is exactly the reason that bitcoin and blockchain works. It's because there's so many copies of it. There's so many copies of the bitcoin ledger that everybody has, that if anybody comes in and tries to change the bitcoin ledger, everybody goes, no, you can't do that. So it seems that there's a great parallel here between the trust that we have in bitcoin and blockchain and the trust that we do in scripture. There's thousands of parchments of scripture, and there's an incredible congruence to them. Okay. You know, Baylor has an oral history program here, so we're trying to do that same sort of thing. So most of the cases that don't have written language usually have an oral tradition. I think this was true definitely before any written language. And this is one of the problems now. For example, I think that we're having with translating the Bible. The thing that AI is really interesting in terms of its application is translation of scriptures. We think of Wycliffe and their organization and how they oversee translation of the Bible, and they have to do this with people with an oral tradition. And this has to be difficult. I was actually surprised at the number of languages that the Bible hasn't been translated into. As somebody in missions, I'm sure you're interested in that. Could you unpack that and tell us a little bit about that? [00:09:37] Speaker B: Yeah, it's interesting. The number is quite large, and it's a very difficult thing to talk about because from an American standpoint, we're very accustomed to thinking about. So it's in English or it's not in English. In a lot of cultures, there's a lot of people who speak multiple languages. And so you may have. Maybe it's not in this tribal language, but it is in this language of wider communication. And I think that is kind of a struggle that we're dealing with today, is that we, particularly those of us who grew up speaking one language and only one language, it's very hard for us to understand what it means to speak 1, 2, 3, 4, 5 or more languages and to be conversant in multiple languages. Maybe this domain is for this language and this domain is for another language. So I only speak English and Spanish. I can defend myself and get away with a couple of other languages. I can kind of make it through context somewhat very short period of time, have some written abilities or be able to read other languages. But it's Very hard for people who only speak one language to think about that. But you could technically have someone who's. The Bible has not been translated into this one minority language, but that same person has access to Scripture in, say, Spanish in a wider communication, and they speak both of those languages. And so there's other languages where it's not just the Bible hasn't been published or been printed in those languages. There really aren't anything published in those languages. They're oral languages only. And so it's not as simple as, well, let's just translate this. Let's just put it in the words. Maybe there's not an Alphabet, maybe there's not a way to even translate because it's just an oral community, an oral spoken language. [00:11:34] Speaker A: So don't. Why does it matter if they can read it in Spanish? Like, if I know Spanish, that's a national language for me, but I also speak a tribal language. Why does it matter if the Bible would be translated into my tribal language? [00:11:49] Speaker B: You know, I like to say I've been speaking Spanish now for 20, going on 29 years. And so I feel like I speak fairly good Spanish. I understand Spanish quite well and, and I think I speak very well. But if I was arrested in another country and put on trial and they said, hey, you speak perfect Spanish, we're going to do the trial in Spanish, or if you'd like, we can give you this, we'll let you do the trial in English every time I would take the English, even though I'm very confident in Spanish, my life would be on the line and I think I would rather have it in the language that I grew up speaking as a child. Well, what more so importance than Scripture and having access to God's word in a way that is understandable in the most clearest sense. And so that's why to me, having this understanding, particularly of the metanarrative of Scripture, the whole story of Scripture, I think that's incredibly important to allow people to, to have access to you. [00:12:57] Speaker C: And I have a common friend, Terry Lassiter, who spent time with his wife Vicki. Terry, he is a dentist. He also teaches my life group at church. But the time that he spent there partially was to do translation. And I was really surprised at the difficulty of this and the time that people have to spend in the field in order to get a translation of the Bible. Yeah. [00:13:20] Speaker B: So I do think that this is one of the really cool breakthroughs that AI has helped with is in the area of translation. And not. It's still funny, whenever somebody Says, hey, I would like a new version of the Bible in this language. Can you just drop it in and give it to me? Well, no, it's a little more difficult than that. You don't just drop in the Bible and say, okay, now give it to me in this other language. And it comes out swimmingly. Well, I think you have to spend a good bit of time doing some testing and checking a significant amount of the time that is done in translation, be it oral. So we do. I did a lot of work in oral Bible storytelling, and we wanted the same output. We wanted clear, accurate, natural, retellable stories. Those were what we were looking for. And a lot of that work is people think you could just, well, I'm just going to translate it really quickly into this other language. You really want to do a lot of testing with community to make sure that people actually understand the community understand the language. You've used the right words, you've used the right the tone. You're really clearly communicating exactly what the Bible is communicating in the written form. And that's a lot more difficult than people realize. [00:14:36] Speaker C: Well, in fact, the original 2017 paper that introduced transformers, which are the backbone of ChatGPT and Grok and Claude and all of these other incredible AI tools that we have today, it was motivated by the idea of translating languages. Yeah. And you just can't. You just can't do it word for word. Germans, for example, usually in their sentences with verbs, and you have to figure out. You have to wait till the end of the sentence in order to figure out what the heck they're going to do. [00:15:04] Speaker B: So Jonathan and I are working on a project right now with, with deaf translation and trying to help this organization that I'm a part of, Deaf Pathway Global, trying to help them be able to translate sign roots so that we could translate the Bible into many different sign languages. And I was surprised at the challenge that they have of they start out their translation with the place and they describe the place and they've got this whole series that they go through for every single passage. And it was very different than the way you and I would look at it and we just translate it. But they have to paint the picture of the place and then the people and then the numbers and how many. I'm getting the order wrong now, but then what they were doing the action, and there's all these things that every single translation has that even if they're just telling the story, I went to the store, they have to start with the store and they have to Start with who and then what they did. And it's so complicated that it's. And we just think, okay, with a subject and a verb and an action are, you know, an object of that. No, it's much more complicated than that. [00:16:13] Speaker A: The entire grammar of visual languages, sign languages, it's. It's set up to help the signer visualize and picture what's happening, which is just amazing to think about that that's built into the grammar. But even the order of which that you'll. The order you use words in the sentence is you're building a picture. And it's really amazing that Transformers are able to encode a lot of that information. Like the success that we've gotten with text translation going between, like English and Chinese, for example, I mean, it's greatly, massively changed localization industries. [00:16:57] Speaker B: And it's in the last. What you were getting two years ago and what you're getting today is absolutely remarkable. And I liken it to two years ago when I was using some of these commercial tools and a little before that, before there were commercial. We're using some other tools. It's like I went back and I'm watching television from the 1980s or 90s with the black things on the sides of the screen. You know what, when we were watching television in the 90s, we thought that was phenomenal transmission of a broadcast. The amazingness of having someone on the other side of the country in a studio being recorded. And I'm watching that on my television live now. The quality is 8k in some cases. And I can. I mean, they have to kind of dumb it down because you see every little wrinkle that the person has. And that's sort of what's happening with AI right now. Where two or three years ago I was blown away with the quality we were getting. And then today I look at it and go, wow, this is realistic. Two years ago, that looked like trash, but at the moment it felt really good. [00:18:07] Speaker A: So for some of the listeners that might be curious, I think it's really interesting to like, one of the things I've done over the last couple months is spend a lot of time researching how people are starting to apply machine translation to Bible translation. And one of the places I started investigating this is how is mainstream localization agencies using this. And I found some research papers talking about, in most cases, for documents that are less critical or just moderately critical translators. They're really just consultants. They're checking the result. And the, you know, these AI tools are producing most of the Actual text that's being. And that's just being checked. When you get into really important things like legal documents or maybe trainings, then the more specific it gets, the more important it is, like medical information. If that's mistranslated, someone could die. And so then you. You need more people in the loop at that point. [00:19:11] Speaker B: It's interesting. You use the translation, the medical. My wife's a nurse practitioner, and they use an AI that does most of the draft charting. And it's pretty amazing that in the past, you would have had a scribe that was walking around, trying to keep up, taking notes, maybe a little. Probably more medical background training, but not officially. Not a medical practitioner, but taking notes. And then my wife would then chart at the end of the day based upon the scribe's notes. Now, the AI, and she could correct me, tell me how I'm messing this up, but my understanding is like the AI is just charting everything that she does, and she has to say it out loud. The AI captures all that. And now she spends a lot less time at the end of the day going back. She still needs. She has to read everything and check everything, do everything, but it is so much faster. And here's the nuance. Almost all of her exams are in Spanish and her charting is in English. And so it is like instantly putting all of that information from all of her interactions with patients and with parents, because there's pediatrics. And then all of that goes into the chart. And so you get much more robust charting. It's amazing. The quality of that is amazing. And the time that it saves, just remarkable. [00:20:34] Speaker A: So that's where we would like to get to with. With Bible translation. And, you know, as. As a Christian who believes that God's word is. Is that it's given by God, it's very important that it's not wrong. It needs to be clear, it needs to be understandable, it needs to be accurate as much as it can be. So I would put it in a critical category like I would a medical document or a legal document. And another thing that's important to realize is in languages with less resources, less written text, the quality of these machine translation tools is much lower. And so that's. Obviously, if you're translating the Bible into a language that doesn't have a translation of Scripture, there's probably not a lot of resources in that language in many cases. And, Don, you can talk about this. I think in some cases, there's not even a written Alphabet. [00:21:27] Speaker B: Right? Yeah. Which makes it. This is why looking more at an oral Bible translation than a written Bible translation in some of those languages. And I think you picked up on something that's important. I think a lot of people have this idea that Bible translators are going out and they're just by themselves translating the whole Bible into this other language. And in reality, there's a team of people who are translating mostly national mother tongue speakers of those languages, and they're doing most of the translation. And then the person that is the translator is really guiding that whole process. So if you think about how AI is accelerating that process, there are some very difficult things to translate in scripture that no matter what language you're going to, this is a very difficult concept to translate. There are other parts of scripture that are much easier to translate. Narratives. You can translate the story a lot easier than some other parts of scripture. So if the AI knows what those difficult parts of scripture are to translate, it can do lots more checking of the translation in those parts. We built this thing when we were with our Deaf pathway group to help them think through what are the different parts of the scripture that are more difficult to translate. And then giving them the AI actually suggest how they might consider translating that in a way that would be a little easier to understand. [00:22:57] Speaker C: So I looked up on ChatGPT, and this is what they told me. So it has to be true. [00:23:01] Speaker B: You can't trust it. [00:23:03] Speaker C: I can't. [00:23:04] Speaker B: I'm just kidding. [00:23:05] Speaker C: Okay. I don't know. Of course I trust chatgpt. We're good friends. It says that the Bible has been Translated into over 3600 languages. About 700 languages have the full Bible, about 1600 have the new Testament, and the rest have portions. This is the statistic that blew my mind. There are still roughly 3,000 plus languages that do not yet have scripture. With all the effort that's gone into translating the Bible, this seems like an astonishing statistic. [00:23:35] Speaker B: Yeah. And a lot of these groups are quite small. There would be some of these people groups may have a few hundred people in those groups. And so you can imagine. So if you have limited resources, which most of us in ministry have very limited resources, and it takes. Up until recently, it has taken a lifetime of someone going and giving their whole life to doing a Bible translation. And usually just you're getting the New Testament. Often, as you cited there, it's very difficult to do that for a people group of 50 or a people group of 90. [00:24:16] Speaker C: Are there really groups that small? [00:24:18] Speaker B: Yeah. [00:24:18] Speaker C: That speak a. Wow. Okay. [00:24:20] Speaker B: Yeah. They're usually a dialect of another language or a tribal group. So there's some that are very small. And then if you look at the scale, like the viability scale of these languages, oftentimes they're dying out. And so there's very few people left who speak that language. The majority language is overtaken those other languages. And so you end up with this precarious situation of, well, it's going to take us 25 years to get this translated, but by the time we finish, nobody will be left who speaks the language. [00:24:53] Speaker A: There are still a lot of minority languages that are still viable and some have millions of people. But this is an important part of when you're deciding how to allocate a little bit of ministry money and a few people who are willing to spend 30 years on this, it starts making sense why these languages don't have a translation of Scripture. [00:25:19] Speaker B: And if the person is really there helping facilitate the translation, it's very difficult without people that speak that language natively. Those really have to be the translators. And so sometimes you'll be going down this road and then that person doesn't want to participate anymore. And you don't need to start over, but you take a significant setback. [00:25:43] Speaker A: One thing I've learned over the last year or so, looking at this, is one of the ways that AI can be really helpful in shortening the amount of time that it takes to do one of these translations is in something that is called a draft zero translation. So like the SIL and the E10 innovation lab have done some really cool stuff with this. But basically my understanding and Don, please add on correct if I'm missing something here, but my understanding is that these groups, like Don mentioned earlier, there's different genres of scripture. So there's poetry, there's narrative, there's others. So they would have a community, a Bible translation team of these mother tongue speakers translate maybe a couple of narrative books of the Bible and then based on that, they can generate a draft of other narrative books of the Bible and then they'll have human editors go in and change and correct. But that even just having that starting point can save a massive amount of, even just the time it would take to type, but it can also save a lot of time in other ways. And so by using AI as a tool, we still have humans in the loop. This is super important, but you can save a ton of time with this. Potentially. [00:27:12] Speaker B: Yeah. So when we would do oral Bible storying, it would be this is kind of a draft zero of the story we called them. And people did not like this, but we Called them kind of a throwaway story. Like, we just want to get a quick draft of what that meta narrative of the story itself. The problem is that's not a, it's not a throwaway because we put a lot of work into those. So they're not like you don't want to offend people by saying, well, everything you've worked toward, we're going to throw that away and start over again. But it is, it gets you somewhere really quick, quickly. And so I'm a very big iterative person. Like, I know that some people like to polish and polish and polish before they ever push anything out. I would much rather get an idea out there and see if it sticks and then we can keep polishing it and get better and better and better. And I think that's kind of a new approach. There have been some who really didn't want anything out there in the wild until we get it really, really super polished, that's really hard because particularly on the oral Bible storying side, we want to tell the story and to see the response and to get the feedback. And do people really understand what we're saying? And if they do, are we saying the right word? Are we using the right word for spirit? Are we using the wrong word for spirit? And so there's, there's something to be said for the iterative approach to a lot of the stuff that I do. AI or Bible story in general. [00:28:44] Speaker C: By the way, listening to the biography of Elon Musk, that's what he does. He does exactly what you're talking about. He puts a product out there that isn't finished and then looks at how it works and its ramifications. And he does fine tuning. There was just a post, I think, of the last three Raptor engines that he used, and they are incredibly improved from one generation to the next. So it's good to put it into practice instead of trying to stay behind the curtain and just doing all the polishing without the interaction with reality. [00:29:15] Speaker B: Well, you have a new Tesla. [00:29:18] Speaker C: I do have a new Tesla. [00:29:19] Speaker B: As a Tesla driver myself, I can tell you that they frequently push updates and it's amazing. Like the whole idea, it's like driving a computer. The whole idea of just, oh, we're going to fix this. And it just is while you sleep, you wake up the next morning and you've got a new feature and this has been turned on or this has been turned off. And so I think that whole approach to quickly be able to iterate and get fixes and get changes out. It's a whole different approach and some people aren't comfortable with it. Other people like myself, that's kind of. I love it. [00:29:54] Speaker C: I think it's astonishing. I thought I knew about self driving cars until my son made me go take a test drive and I am just blown away. It's the best engineered technology since maybe the cell phone. It's just astonishing what the self driving Tesla can do. Let's get back to the translation of languages. I'm sure a lot of these cultures don't have a written language. What do you do then? Do you leave them an oral representation of scripture or how does that work? [00:30:23] Speaker B: I'm a big proponent of orality and using the modes of learning that the people are accustomed to. And so I'm a big proponent of helping people understand metanarrative through the whole Bible, telling all the stories and helping people understand that if people don't have a written language, I'm not opposed. I'm all for everybody getting a copy of not one version of the Bible, but maybe three or four versions. Three or four. A dynamic equivalent. A one that's very word for word illiterate. I mean there's all sorts of needs out there. Again, we're coming back to how do you prioritize when you have limited resources and oftentimes it is the case that if we can give them a really good oral storied version of the Bible and then they also are very fluent in another language, then maybe they could access God's Word in the trade language. And here's the reality, I think, and I think we've seen this, other orgs have seen this where you have people excited about God's word, then all of a sudden they have the desire to see God's word translated into that other language. In a project we were doing in Mexico with a group story together is what we had. We were training some people that were in another language that didn't have the Bible in their language. They were using Spanish as the trade language. People were not getting everything. And so we did a storying project, helped them understand the scripture in their own language. And now all of a sudden there's this renewed effort from within their own people group to say, hey, we want the Bible translated into our language. And so now the drive is coming from within in the people group, not somebody on the outside saying you should have this translated. But that starts by having a good oral understanding of God's word. [00:32:25] Speaker C: One of the things that we're debating now in the United States is the Influx of different, if you will, cultures. I would imagine this would be a big barrier to translating scriptures. You have to be theologically sensitive. You have to be aware of the political and social barriers, and this has to be different. I think it's one thing to actually translate scripture, it's the other one. Another problem is to getting it to [00:32:51] Speaker B: be used, I think. Yeah, and what you don't want to do is you don't want to compromise scripture just to get someone to accept it. So we don't change scripture because it's more palatable, but at the same time, you don't go out of your way to offend people by just not contextualizing. So contextualization, sometimes in the past, I think it's gotten kind of a bad rap. I think you can be so contextualized that you become inaccurate because you can be so uncontextualized that it is inaccurate. And then you can overly contextualize. And so I think the contextualization is trying to make God's word understood by the people who are hearing it or reading it. And I think that's very important. And so it's not just translating word for word, because, again, in America, we primarily have grown up speaking one language, and it's very hard for us to understand that words have more than one meaning and words have multiple meanings, or There may be 30 words for this one concept. And picking the right word is very important. [00:34:01] Speaker A: Don, can you give an example of what contextualization might look like? [00:34:07] Speaker B: Yes, I can. [00:34:09] Speaker C: Would you? [00:34:10] Speaker A: I just think that might. That might help our listeners kind of have a picture of what this might look like. [00:34:16] Speaker B: I think just the whole idea of a. Trying to think without saying where this was. But there was a certain country in North Africa that we were working in. And the reality is, every one of those groups that we were working with, there are lots of different tribal groups. All of them are very oral learners. And a contextualization for. There was actually the Bible stories themselves, to take God's word, be very faithful to God's word, and instead of just saying, hey, here's a bunch of books, you can go read these. Ironically, I had heard all of my life the story of people who have translated the Bible. And there's warehouses full of Bibles that nobody can read. And I literally was in this North African country, hundreds of miles away from electricity or cell phone service in the middle of nowhere. And there was this little warehouse went in there, and it literally was full of Bibles that bugs had eaten through that people were not Reading because they couldn't read and they weren't interested. And I'd heard the story all my life and I actually saw it with my own eyes that, you know, pick up a box and the bugs went everywhere. They'd eaten through the scriptures. And I think the contextualization in that point was we're actually taking God's word and we're putting it in an oral format so that the people would actually hear those words and they would understand the words. And these were not Christians that we were working with. We were teaching them how to story God's word in their different languages. We had several different tribal languages. And the hope is that coming out of that, they would actually want to read God's word. But the contextualization in that case was, we are going to tell you the stories. We are going to teach you the stories. We want you to teach other people these stories in hopes that God's word would be planted in the hearts of the people in that region of the world. [00:36:18] Speaker C: It does seem like artificial intelligence and transformers are going to have a big impact in this area of translation. To what degree has Wycliffe and your organization, the International Missions Board, incorporated AI? Is this something which is going on full being full blown, or is it something which is just transitioning what's going on? [00:36:40] Speaker B: I think SIL or Wycliffe has done, obviously been cutting edge in using AI for different things, particularly for checking, testing, and looking for problematic parts of scripture. It can really speed up the process. It can help the people who are doing the translation can queue up. They've got this model for oral Bible translation, which is a translation, not a story. But the oral translation is. They've used some tools that can help speed that process up. We use it a lot in a lot of other areas, so we use it for translation. I also use not Bible translation, but just translations in general. Like, we need this resource in this other language perhaps. I'm going to try to think of what I can share, what not to share. But there's some meetings that I'm having right now with lots of people, and there's a lot of people on those meetings. And they're long meetings and we have hundreds of hours of content that we're gathering feedback on. And so I've been able to build some solutions that will just digest all, all of that information and then pick up patterns and do some sentiment analysis on how people are feeling about different things. We've used it for some of our. We do a lot of digital engagement, so we build out basically an ad campaign like you would if you were selling Coca Cola. Except we are understanding. We're helping people understand more about Jesus. And so we'll run an ad on social media platforms in a certain language in a certain part of the world. And the call to action, instead of going and buying Coca Cola, may be chatting with us and asking spiritual questions. And last year we had around 80,000 people. We had millions of impressions. Those are not really important. What is important is we had over 4 million people download or watch an entire evangelistic video. We had around 80,000 people send us messages. So those messages, that's great. The problem is you got to have somebody on the other side to respond to those messages. And so a lot of the people, a number of the messages come in are spam, are there porn, are there people yelling at us? They're not really wanting a spiritual conversation. So we've been able to take the AI and this is kind of really cool to see us leaning more into this. But the AI can ascertain quickly as whether or not this has potential to be a spiritual conversation or is this just a hater that's out there? So if it has the potential to be a spiritual conversation, those get prioritized to a human to chat with them. So it's like going into the ER and saying, hey, I've got a splinter and I've got a guy over here with a pulmonary embolism. Then the nurse, the triage nurse is going to take the pulmonary embolism and put them to the front. And the splinter may spend a few hours in the er. Well, the same thing for our responders, our human responders. We want to send them people who are really wanting to have gospel conversations, spiritual conversations. And so the AI can help us triage, if you will, that backlog of conversations as they come in. So we created faithbot through that. [00:40:09] Speaker A: So, Don, I want to get to faithbot. One more thing. Just thinking, kind of closing up this discussion. Dr. Marks, you asked about kind of what people had been doing in the Bible translation space. [00:40:20] Speaker C: Sure. [00:40:21] Speaker A: There's a really interesting article, if our listeners are interested. It's called. It's E10, just the letter E10. Innovation Labs Are AI drafts really faster? And this article goes through, I think it's like 75 projects that they've been doing, applying these things to. It's got some really interesting data. So if you're interested in this, there are people working on it, doing some really good work. [00:40:47] Speaker C: Jonathan, send me that link and we'll put it in the podcast notes. [00:40:50] Speaker A: Okay, yeah, absolutely. And one more thing, just to. Just to kind of paint a vision for this, even if you're not a Christian, this is really valuable because these same. You. You have these same paradigms with other things you may have. People want to have medical records in their tribal language, so that's easier for them to understand. I. There's a lot of really incredible, very smart international grad students that I have the pleasure of working with at Baylor. And it's interesting. Many of them speak technical English way better than they speak conversational English. And so when I'm talking to them over dinner, they might not understand what I'm saying. But if I read a paper that most Americans have no idea what I'm saying because it's full of technical language, they get it perfectly. And you can actually use Bible translations as a seed translation to. To enable the rest of the Internet, the rest of the knowledge of humankind to be translated into their language. And so as a Christian, I think translating the Bible is the most important thing. But even if you disagree with me about that, there's a ton of common ground. There's just so much you can do with these translations from a just social, good, good of humanity perspective. So it's really exciting space to be working in. [00:42:12] Speaker C: That's really fascinating before we get into faith, but I do want to talk a little bit more about signing. I have learned so much, mostly from Jonathan and watching some videos about signing. It's incredible. I've learned, for example, that if you are deaf since birth and you learn sign language, and there's another person who has learned sign language later in life, but they are able to hear that. The person that is able to hear when they learn sign language, the deaf person could actually tell they have an accent. So they have an accent in translation. And the subtleties of the sign language are awesome. We were in the presence of that, Jonathan, and I was just blown away by the beauty of the signing. The sign language is excellent, yet the ability to translate sign language to text or sign language to oral or the backwards problem is really in its infancy. We saw some cases where these AI robots were doing things, but they were clunky and the deaf community doesn't like them at all. So you've been looking at the ability of AI to make this signing more accessible to the deaf people, and that would be a really big boon, I believe. Could you unpack that a little bit, Jonathan? [00:43:35] Speaker A: Yeah, sure. So it's interesting looking at the history of AI and different applications, I Mean artificial intelligence. It's been around a long time. [00:43:43] Speaker C: Oh, yeah, Just like me. [00:43:46] Speaker A: Yeah. I think it's the Dartmouth conference in the 60s, kind of where a lot of this started. It's really interesting to see. There's just been these different algorithms that have come out that have enabled new applications. And one of the things that we've kind of just recently gotten to is the great success that we're seeing now with machine translation. And that's been enabled by Nvidia making such wonderful GPUs for us. Thank you. And also just all of the training data we have on the Internet. So if you can go back to the. [00:44:24] Speaker C: Okay, for those who are not nerds, what's a gpu? [00:44:27] Speaker A: That's a graphical processing unit or video games. [00:44:30] Speaker B: Right. [00:44:32] Speaker A: That's one great use for them as well. [00:44:34] Speaker C: That's what they were initially made for. And they have applications in AI and they're so cool. [00:44:40] Speaker A: But yeah, I mean, the math for this, at least some of it, at least the idea, knowing it's possible has been around a long time. Claude Shannon has papers on this from 1950 about predicting the English language. And he, at the time, he gave up because the amount of human labor to generate the statistical predictions, just astronomical. But what's changed over the last 70 years is we suddenly have most of human knowledge, at least in English, on the Internet. It's already in a format where it can be used to train an algorithm like this. And we have these GPUs. I'm a big fan of space and the Apollo missions and Artemis. And I mean, it's just amazing to think the amount of computing power in the Saturn 5 rocket is just. My phone has like millions of times more compute. [00:45:46] Speaker C: Yeah. [00:45:47] Speaker A: And that, that level of just. I mean, it's orders and orders and orders of magnitude more. And just what we've seen with Moore's Law, that's enabled us to get here. [00:45:58] Speaker B: Yeah. [00:45:59] Speaker A: And that's letting us get now with machine translation to accuracies that can sometimes be like high 90s. It's why localization can use these machine translation techniques and then just have a human go and review it. But with sign translation, we're not there yet. It's a much more difficult problem. If you think about there's not videos, there's not abundant training data. Most people don't realize this. There's actually hundreds of different sign languages that are distinct languages. And then how do you tokenize signlink? I mean, there's lots of people doing great work on this and the field is making lots of advances, natural language processing. It's also got computer vision angles to it. But there's some really interesting work. If our listeners are interested sign mt, you can go to their website, it'll be in the show notes and see that. So there's some, I mean if you as someone from outside the deaf community looking at it, it looks good. But that's one of the real dangers in this research space is it's easy, it's, we're not, we're not getting high enough quality. There's a really important study that I found that evaluated some of the frontier real time models for this and it found that they were getting an deaf person was able to understand about 30% of what was actually being translated. I mean, that's awful. [00:47:30] Speaker B: That's awful. [00:47:31] Speaker A: That's like getting some of the words right. That's not good enough to use for a phone call that I don't care about that much. It's certainly not good enough for really important things. And it's also important for us to realize that there's a lot of, we have to be very careful as engineers, computer scientists, PhDs from outside the deaf community coming in. One of the things that is so helpful about the collaboration that we're building is we have a lot of deaf partners that are active in the research and helping us understand what needs to be done. Because they're the ones who own this, this is their community and we get the privilege of working with them, but they're the ones who can identify like where is this helpful? But yeah, 30% versus something like high 90, maybe 98% for some of these machine translations for languages that we have a lot of training data, we'd like to get to that point where you could just put in text and then get a sign language video out or maybe go from a sign language video in one language to a sign language video in another, another language. We're not there yet. [00:48:45] Speaker B: Yeah, that's kind of when the three of us were at dinner one night and I think Dr. Marks asked, so what would you like to see AI do in ministry that it's not currently able to do? And without a doubt, without hesitation, I'm like, well, this is the problem in innovation space. I always work on lots and lots of innovative ideas at once. And then I always have a moonshot that's out there that's like, this is impossible. I don't see a way to ever do this. And this is my moonshot. I would love to see this done. It's Going to take people a lot smarter than me. That's why I'm on the call with you guys. But I think that this is. I was encouraged by. You all talked about this. Well, yeah, it's just math. I've used that a lot. When Dr. Mark said, oh, it's just math, I'm like, yeah, but it's like really hard math. [00:49:32] Speaker C: It's hard math. [00:49:33] Speaker B: It's really hard math. I'm excited about the possibility of even attempting to create something like this. And there's a lot of people working on it. But I think a lot of people, like Jonathan said, they don't understand. They just see. Well, it's just hands, and it's not just hands, and it's. It's proximity, it's leaning in, it's leaning out. It's emotions, it's facial expressions, it's how far the hands are. There's all of these things that are happening at the same time that for a deaf person, makes. It's just intuitive how that communication is happening. And from an outsider, it's very hard to see. And I've had a lot of people just tell me over and over, well, I mean, how hard would it be to translate the hands? I'm like, well, it's not just the hands. Way more than the hands. [00:50:22] Speaker C: In fact, the signing can be so subtle. For example, a little eye twitch. Twitch means, I understand, I'm following you, I understand what you're saying. And the expressions on the face and the subtleties of the movement, that's what I think is going to be the difficult thing to translate. I think the big hand motions are going to be easier, but the little subtleties of the face, because we're born to recognize faces, because we got to know mommy's face and daddy's face. And so we have this great ability to interpret facial expressions. And that's certainly used in sign language with the subtleties of the. Of the expressions. But I tell you, it's so beautiful to watch. Yeah, signing is beautiful to watch. [00:51:06] Speaker A: I mean, it's one of the hardest problems I can think of. Like you said, Don, it is so hard, the amount of compute that you need to even deal with this as a, like, you know, it's a computer vision problem. It's hard. I think. I do think it's a problem we can solve as a species. Like, I think it's. We can. It's possible. [00:51:30] Speaker B: I thought you were talking about the three of us, and I was like, you're going to have to pull my weight. [00:51:34] Speaker A: I would love to. I would love. Yeah. I mean, there's. I think there's a lot of really smart people working on this all over the place. And, you know, as a scientific community, we want to contribute towards building a solution at some point. I think it's. It could be. It could be a. It could be decades, possibly. But it's. I mean, it's a. It's a problem that if tools were developed by deaf communities and. And just with input, it could. It could potentially really help a lot of people and allow the Bible to be translated, too, which would be fantastic. [00:52:09] Speaker B: Yeah. All right. Can I talk about some of the tools that we've built going back to an earlier question? [00:52:14] Speaker C: Sure. [00:52:15] Speaker B: Go. [00:52:15] Speaker C: Yeah. [00:52:16] Speaker B: So one of the things that the team at the imb, the International Mission Board, we have a small innovation team. There's a couple of us in the states and about 6 of us overseas on my team, and we've created just some really cool tools that we like to share with the church and share with Christians. And the first one we created actually came out of the idea of how do we handle these online conversations with people? And that first one we developed was something called faithbot. Just faithbot. [00:52:49] Speaker C: IO, yes. [00:52:51] Speaker B: It's very simple. I know you've used it, you've tried to break it, you've tried to make it. See if it could answer correctly. [00:52:58] Speaker C: Before you go on, I want to tell you my concern about faithbot. And it's been answered. And I want to do the concern with a story by Emo Phillips. Now, Emo Phillips is the greatest comedian of the 20th century. In fact, I went and saw him in Dallas last week, and here is the this is a story he tells. And it was voted the 44th funniest joke of all time by Gentleman's Quarterly magazine, and it was voted the funniest religion joke of all time. Okay, you ready? [00:53:27] Speaker A: All right. [00:53:28] Speaker C: Once I saw this guy on a bridge about to jump. I said, don't do it. He said, nobody loves me. I said, God loves you. Do you believe in God? He said, yes. Are you a Christian or a Jew? He said, a Christian. I said, me, too. Protestant or Catholic? He said, protestant. I said, me, too. What franchise? He said, baptist. Me too. Northern Baptist or Southern Baptist? He said, northern Baptist. I said, me, too. Northern Conservative Baptist or Northern liberal? Baptist. He said, northern Conservative Baptist. I said, me, too. Northern Conservative Baptist, Great Lakes region or Northern Conservative Baptist, Eastern region? He said, Northwest Conservative Baptist, Great Lakes region? I said, me, too. Northern Conservative Baptist, Great Lakes region. Council of 1879 or the Northern Conservative Baptist Great Lakes Region Council of 1912. He said, Northern Region Baptist Great Lakes Region of Council of 1912. So I yelled, die heretic. And I pushed them off the bridge. Why that is so funny is there's so many nuances between the differences of evangelical Christianities between different denominations that I think I thought it would be very hard to write a chatbot, which is what you've done. And by the way, it's marvelous and everybody needs to go and play with it again, as Don mentioned, it's Faithbot IO Faithbot IO and go there and you can ask your questions just like you do with chatgpt Groker Claude. And I tell you, I was really impressed, Don, with the, with the responses. Now, you're the father of this. You're the one that put it together. You tuned it and babied it and made sure it was good. How did you keep from, from avoiding this, this denominational challenge between different denomination? [00:55:29] Speaker B: Yeah. So the cool thing is I. At the very top level, you're grounded in scripture. And so it specifically did not feed it a bunch of these tertiary issues that then you always say this. You always say this. So instead of like painting it to be exactly like an echo chamber of maybe me, it is trained to always respond as an evangelical Christian would respond. But the caveat is it's supposed to ground it with scripture and give you a reference saying, well, this is where this is found. So if you force the model to always respond with scripture, it makes it really hard to go outside of scripture. I know that sounds simple, but. And I would say that we tried a lot of approaches. I wish I would have tried this. The first one, the first approach we had. I'd gathered thousands of these historical documents and we tried to do like a rag where all this information is here. And the reality is we found that I know you don't like this word, but it hallucinated a lot. And the reason is we were very explicit on all of these topics. But if you weren't explicit, then it just went and it created something. Or sometimes it may have multiple contexts of the same issue, but dealt with differently. So what I found has been the best is less is best. And we just kind of fine tune it down to exactly what it is that an evangelical Christian would believe, but then grounded in Scripture. So that's. That's the first one we built and that now has. Just over 800,000 people have used that since we released it April of last year. So it's been a Little over a year. We've just had over 800,000 in English, and then we have multiple versions of it in other languages. So if you look at those other languages, the numbers are significantly higher then coming out of that. I would say that if you're wanting to share a tool with someone who's just maybe a seeker, someone who's got questions about faith and they're looking for answers, that's a really good tool to share with them. I like that. And as I'm traveling overseas, I find that when you're sharing with people, frequently in the Middle east, someone will open up their phone and they will go to ChatGPT to respond to your answers about faith. [00:58:00] Speaker C: Yes. [00:58:01] Speaker B: And so what I would like to do is ground those answers so that because ChatGPT may give you a good answer or it may not give you a good answer, and this will always give you a very good answer. The other tools that we've built. So if you go to faithbot Tools, so IO is this thing you can share with seekers, then Tools is aimed at believers who want to better prepare themselves for how to share share with other people. And so this is where I would send most people who are listening probably to your show is go to faithbot Tools. You can pick a worldview. There's a button called Engage Lostness. You can say, hey, I want to share with this worldview. And it will evaluate your ability to share your faith with that worldview. It takes on the Persona of that worldview. You can carry on a conversation back and forth, and at the end, it'll give you some pointers for, hey, you said this. But when you say this, this is how a postmodern interprets what you said. You thought you were communicating clearly, but what they heard was this. And so that's about 20, 25 different worldviews that you can choose and you can practice sharing. There's several things on that website that I think your listeners would enjoy using. [00:59:20] Speaker C: I tell you, I was really impressed with some of the answers I asked it. For example, a fundamental question. What does it mean to be a Christian? Define a Christian. That was the question that I asked it, and it gave me basically a CS Lewis sort of response that to be a Christian means to be a follower of Christ. At its core, it signifies an individual has recognized their sinful nature, acknowledged their need for a savior, has placed their faith and trust in Jesus Christ alone for salvation. And this involves repenting of sin and believing in the death, burial, and resurrection of Jesus Christ. And then, like you said, it gives Scripture verses. So it's not elaborating on doctrine that is external from Christian, from Christianity, from Scripture. And that I think is really nice. And then another one I ask, it started to get rough. The Church of Christ believes you have to be water baptized in order to be saved. So I asked your faithbot, do you have to be water baptized to be a Christian? And it says, this is a response. While baptism is considered by Jesus as an important step of obedience for all who have trusted in him, it is a consequence of salvation, not a prerequisite. And then he gives some Scripture verses for that. So I like the idea that it always refers back to Scripture. So you can go back to Scripture and you could look at that scripture and see what it says and make up your own mind in terms of what the Bible says. [01:00:46] Speaker B: I wish I'd have known you would do all this. I would have just let you do all the testing for us. [01:00:49] Speaker C: You've given it all the hardship. So I ask, what is true? Predestination or free will? This goes back to the Emo Phillips joke, right? All of the Calvinists say, well, you know, everything's predestined. And all the Armenians say, no, we have free will. And it gave a good answer. It says, they're both true, which I believe, I think in terms of the perspective of God, that we are predestined. In our perspective, we have free will. So I think that they're not contradictory. And then I ask it a rough one, which is true premillennialism or post millennialism? This is something that you get all the time and it goes through. And it says, well, yeah, we can talk about this. And here's Scripture to do it. Revelation 21:6. But he said, basically the idea is that that's not important. What is important, what is central is the certainty of Christ's return, of the establishment of his eternal kingdom. So I love this Don because it really speaks authority and it returns always to the Scriptures. So I'm a big fan of faithbot. [01:01:55] Speaker B: Well, thank you. That means a lot coming from you. I'm glad that. Glad we passed. I was a little worried what you were going to say there. Yeah, I would really encourage you to do the faithbot tools. Look at all the resources that are there. These are available for free to people to use and to test out and really have spent a lot of time and effort to create some tools that we think would be good alternatives to what I think is unethical use of AI, which I've been around enough people over the last couple years. And I've spoken to a lot of pastors conferences and done a lot of training. Just did one this morning for a group of pastors in Alabama. And the more I talk to people recently, a leader from Mississippi said the more and more he's talked to pastors, he's fearing that a lot of people are just using ChatGPT to write their sermons. And I would tell you that there are things that we can use AI to do and there's things that we should not use AI to do. And I think that the soul care, things that deal with soul. I think that souls care for other souls. Algorithms do not care for souls. And so when it comes to counseling and it comes to trauma healing and it comes to praying for people and it comes to writing sermons and hearing the Holy Spirit, these are not things that we turn to AI to do. We use AI to help us free up time so that we can do the soul care. So if there's an administrative burden, there's some reports that you have to write. I don't care if that's written by an AI. But what I do care about is you've spent the time in the Word and you've studied and you've sought the word of God for your group that you're sharing with for your sermon. And so we've been building out some tools. They're not ready to release yet, but some tools to help pastors use AI to do sermon research, not write sermons. So it can help you with the exegesis of scripture, it can help you with meanings of words, it can help you find other resources, but it's not going to write your sermon for you. That's really for you to work through prayer and seeking the understanding of the Holy Spirit to do so that using it for trauma counseling. I've heard people want to use AI to become the counselor for people who are dealing with trauma. And I just really push back because you can do a lot of damage to that. And so we're not one. I don't think the technology is there yet. Even if I didn't have a theological opposition to this, I don't think that the technology can do consistently provide because you're not looking for a 96% accuracy, you're looking for 100%. You don't want it giving bad information. So I don't think it's there. But I also don't think that's where we need to be leaning into using AI. We should use it for the Other stuff to free up our time to be able to spend with people. [01:04:56] Speaker C: Excellent, excellent. Okay, well, we're running out of time. Any final words, Jonathan, Don, that you like to say about what's going on? We've talked about. Let's see. We've talked about sign language. We've talked about faithbot IO, which you need to go to. We've talked about. Man, I still can't say it, Don. [01:05:14] Speaker B: Orality, orality, not morality. [01:05:16] Speaker C: Okay, I got it, I got it. And you know, the Bible translation and all of the ways that AI is assisting in Christian mission work, and this has been productive. I've learned a lot. So any final words? [01:05:29] Speaker A: I'll just say Don, I think there's a lot of work that needs to be done thinking about where AI should be applied and particularly for Christians, like you're saying. So I'm really encouraged to hear you just thinking through that. And I think there's a lot of stuff to be done there. One thing that I'm seeing a lot is that a lot of people in Gen Z really are turned off by the digital world. They want a lot of young guys talk a lot about embodiment and just what it means that God created people in bodies. And I think it's going to be interesting to see how these tools, they can be used. They're very valuable, but I could see a cultural pull, pushback kind of based on that. And so I want to ask you, I think these tools can still be used, but maybe with care. Curious how you would think about that. And then how would you deal with someone who said, like, well, you know, faithbot, man, it would be awesome if faithbot is used by God to draw people to Christ. But what would you do with someone who would say, man, I think Christians should be the ones actually sharing the gospel, or how do we connect this young person asking questions with an actual person? What would you say to that? [01:06:57] Speaker B: Yeah, I would agree with that. I think that that is a plan that we should be following. We should be telling people about Jesus. At the same time, I don't want to limit any tools that we've been able to create through God's abilities that he's given us. So AI is. This is going to sound a little odd, but AI is just a tool. And AI is a tool, just like a hammer is a tool, just like a pencil's a tool. And so you can use those for good or you can use them for bad. I don't want to throw out AI just because right now it's being used, driven by people who are trying to make trillions of dollars. That seems to be the driving focus of every bit of AI development or to destroy another country. That's kind of the. Those are the things that are driving the AI business right now. Christians are just absent from this. And there are Christians speaking into it, but nowhere near. I hear a lot more Christians are very negative towards, towards AI less today than a year or so ago. But still, there's enough out there. And my response to those folks who are very negative to this is that one, it's a tool. Two, Christians have always been very innovative and throughout centuries, modern medicine, greatly influenced by Christianity. Modern farming techniques, greatly influenced by Christianity. Communications, the history of communication, printing press, Morse code, shortwave radio, broadcast, all of those things have been greatly influenced by Christians who have been in those fields and using those tools for the furtherment of God's will and God's work. And so there's always been people opposed to it. There's always been the person who said, well, you know, you should never preach on the radio because you should. Only, I can't believe that. But I went and looked it up and researched it. There are people who fought the ability for someone to preach on the radio. Nope, it has to be in person. I'm like, I don't find that in scripture anywhere. And so I see it as just another tool that God has given us to be able to reach millions and billions of people that we've never been able to access. And right now, there are places on earth that I cannot get to. I cannot speak those languages. And if God would use this tool, if he chooses this tool to spread his word amongst people that I don't have access to, who am I to say, nope, God, you're not allowed to do that. [01:09:42] Speaker A: Great answer. [01:09:43] Speaker C: Great. Well, thank you. Thank you, Don. This has been wonderful. [01:09:47] Speaker B: Thank you. [01:09:48] Speaker C: Jonathan Swindell and I have been talking with Dr. Don Barger. He is the Director of Innovation and Artificial Intelligence at the International Mission Board. And if you want to visit their site, it's imb.org that's imb.org and so until next time on Mind Matters News, be of good cheer. [01:10:18] Speaker A: This has been Mind Matters News with your host, Robert J. Marks. Explore more at MindMatters AI. That's MindMatters AI. Mind Matters News is directed and edited by Austin Egbert. The opinions expressed on this program are [01:10:40] Speaker B: solely those of the speakers. [01:10:42] Speaker A: Mind Matters News is produced and copyrighted by the Walter Brattle center for Natural and Artificial Intelligence at Discovery Institute.