Latest / Centrally Speaking / Christians and Artificial Intelligence

Transcript

Christians and Artificial Intelligence · Centrally Speaking

• 0:01
  SPEAKER_00: Centrally speaking, it's the Central Schweinfeller Church's podcast.
• 0:05
  SPEAKER_00: It speaks about issues that would be of interest to our society.
• 0:09
  SPEAKER_00: In particular, it addresses how a Christian worldview intersects with Western secular culture.
• 0:15
  SPEAKER_00: In the spirit of the church's founder, we take the perspective of the middle way, which is in agreement with the historic Christian Church.
• 0:24
  SPEAKER_02: I'm Dr.
• 0:25
  SPEAKER_02: Drake Williams, Minister of Mission and Theology at the Church.
• 0:28
  SPEAKER_02: Our website is www.cscfamily.org.
• 0:33
  SPEAKER_02: Today we're going to speak about artificial intelligence.
• 0:36
  SPEAKER_02: Artificial intelligence is the ability of a computer program or a machine to think and learn.
• 0:41
  SPEAKER_02: It is also a field of study which tries to make computers smart.
• 0:45
  SPEAKER_02: They work on their own without being encoded with commands.
• 0:48
  SPEAKER_02: The applications are multiple as this field is influencing healthcare, the automotive industry, finances, cybersecurity, government, law, the military, and even video games, just to name a few.
• 1:01
  SPEAKER_02: Today I'm joined by Dr.
• 1:03
  SPEAKER_02: Michael Marabito, who is assistant professor of computer science at Eastern University.
• 1:08
  SPEAKER_02: He did his PhD in mechanical engineering from Lehigh University in Bethlehem, Pennsylvania in 2019, and a Bachelor of Arts in Mathematics from Eastern University in St.
• 1:18
  SPEAKER_02: David's, Pennsylvania in 2015.
• 1:21
  SPEAKER_02: He is also conducting research in Brownian dynamics of soft matter systems, computational biophysics, membrane-based gas separation, and water desalination module design.
• 1:32
  SPEAKER_02: Michael, it's a pleasure to have you at the show.
• 1:34
  SPEAKER_01: Thanks for having me on.
• 1:35
  SPEAKER_02: Why don't we start with uh a definition of artificial intelligence?
• 1:39
  SPEAKER_02: Um can you um provide one for our listeners today?
• 1:43
  SPEAKER_01: Yeah, I I I like the definition that you gave generally.
• 1:46
  SPEAKER_01: Um so firstly, I think that what's important to know is that artificial intelligence is a broad term that encompasses many different techniques, uh, but generally it can be thought of as any sort of uh artificial computational system that shows intelligent behavior, like you said.
• 2:03
  SPEAKER_01: Um so you know, just complex behavior that is conducive towards reaching goals.
• 2:08
  SPEAKER_01: Also, I think it's important to make the distinction that there are, generally speaking, two types of AI.
• 2:14
  SPEAKER_01: So there's a narrow AI and an artificial general intelligence, right?
• 2:19
  SPEAKER_01: So this idea of a narrow AI, uh, well, that's that's something that has limited abilities, right?
• 2:25
  SPEAKER_01: So this could be just a software system or or a software system combined with hardware to form some sort of autonomous system.
• 2:34
  SPEAKER_01: Um but but this narrow AI is really good, or it excels at a particular problem solving or reasoning task, right?
• 2:43
  SPEAKER_01: So something that's narrowly uh skilled, something like Siri or Alexa that we're well familiar with.
• 2:50
  SPEAKER_01: Now, the other sort of general uh artificial intelligence, this is this is more of a hypothetical uh machine that has the capacity to learn any intellectual task that uh a human being could do.
• 3:02
  SPEAKER_01: And and where we are now, we're still years and if not decades off away from that sort of uh artificial general intelligence.
• 3:10
  SPEAKER_01: But it's it's important to understand those two distinctions.
• 3:12
  SPEAKER_01: And and in terms of narrow AI, that's the here and now, right?
• 3:16
  SPEAKER_01: We we see things, some examples like I mentioned Siri, Alexa, Tesla, Roomba, uh, Netflix, Facebook, these are these are all uh companies and technologies that are very much implementing techniques uh that would fall into the category of narrow AI.
• 3:34
  SPEAKER_02: Okay.
• 3:35
  SPEAKER_02: So you're then saying that if with these two definitions, at least a narrow um artificial intelligence, so most of us have already been exposed uh to this and uh on a daily basis if we're let's say Facebook users or uh watch Netflix.
• 3:50
  SPEAKER_01: Absolutely, and and let's let's unpack this a little bit more here.
• 3:53
  SPEAKER_01: So uh before I I go into a specific example, let me give perhaps a more uh specific definition.
• 4:00
  SPEAKER_01: So so artificial intelligence is the study of an agent that is able to receive percepts from an environment and execute appropriate actions that would otherwise require some level of human intelligence.
• 4:15
  SPEAKER_01: So let me let me put that in different words here.
• 4:18
  SPEAKER_01: So AI is the study of an intelligent machine, right?
• 4:22
  SPEAKER_01: And again, it could be software or could be software married to hardware.
• 4:26
  SPEAKER_01: And this system is able to sense uh stimuli or or data within its environment.
• 4:33
  SPEAKER_01: And that and that's that may be subtle, but that's important to understand that these systems exist within an environment and may give you different results in a, you know, the same question, if you will, may give different results in one environment from another.
• 4:46
  SPEAKER_01: But based on the the input data that this system is able to sense, it's able to execute appropriate actions that require some degree of human intelligence.
• 4:56
  SPEAKER_01: So let's look at Siri for the example.
• 4:59
  SPEAKER_01: All right, and and I would make the claim that there are in general, there are three tasks that Siri has to execute in order to uh have a successful interaction with Siri.
• 5:09
  SPEAKER_01: So you pull out your uh Siri, of course, is is the um the voice recognition software as well as natural language processing uh software that comes standard on on iPhones.
• 5:21
  SPEAKER_01: So you pull out your your iPhone and you you press the home button, and Siri asks you, you know, what can I help you with?
• 5:28
  SPEAKER_01: And um so the first thing that Siri has to do is it has to listen to you and take that audio signal and turn it into text.
• 5:37
  SPEAKER_01: And and it turns out that this is not actually a super challenging problem.
• 5:41
  SPEAKER_01: It it is a difficult engineering problem, but it's not a super it's it's not an AI problem, it's not a computer science problem.
• 5:48
  SPEAKER_01: And and actually, it can kind of be simply done using some mathematical transforms that essentially take a uh use pattern matching, and it takes an audio wave and and simply matches the pattern of that audio wave to a word.
• 6:02
  SPEAKER_01: Okay, so so that's the first thing.
• 6:03
  SPEAKER_01: It is that it has to turn your your audio into textual data.
• 6:09
  SPEAKER_01: And let's go to the third the third task that Siri needs to execute.
• 6:13
  SPEAKER_01: So Siri needs to have a a large knowledge base, a database of just facts.
• 6:19
  SPEAKER_01: You know, uh who won the Phillies game on Thursday?
• 6:23
  SPEAKER_01: How many inches away is the moon from Earth?
• 6:27
  SPEAKER_01: Uh what is the temperature outside right now, right?
• 6:30
  SPEAKER_01: All of these facts that are objective and have answers, Siri needs to have access to that in order to supply you with answers to questions that you may have, right?
• 6:41
  SPEAKER_01: Or or it has to have databases of songs to play for you, or so on and so forth.
• 6:46
  SPEAKER_01: Um and this is also not the challenging part, right?
• 6:50
  SPEAKER_01: This is you know, it may be an engineering problem to build such a database of that sort, but it's not an AI problem.
• 6:56
  SPEAKER_01: The second task that has to be completed by Siri, now that's that's where the interesting things happen, right?
• 7:02
  SPEAKER_01: So after step one, what you've got is you've got the user's question or the user input in textual form.
• 7:10
  SPEAKER_01: And before you're able to answer that user's question with you know some of the facts that they're requesting, you have to take that syntax, those words that have been taken from the user, and you have to extract meaning from them.
• 7:25
  SPEAKER_01: You have to extract the semantics of what the user is asking.
• 7:30
  SPEAKER_01: And this is the interesting part of the problem.
• 7:32
  SPEAKER_01: This is what artificial intelligence is is trying to do, is it's trying to extract semantics from syntax.
• 7:39
  SPEAKER_01: It's trying to extract the meaning of what we say or the meaning of what we write from the form of how we write those things.
• 7:48
  SPEAKER_01: Um and so this is this is somehow gets at the idea of what it what it means to be human and and is is that notion of of abstraction and under being able to understand the meaning of of of the things that we say and write.
• 8:03
  SPEAKER_01: Um and so to be able to train a computer system or an artificially intelligent system to do that, um that that's the discussion.
• 8:12
  SPEAKER_01: That's what we're talking about here.
• 8:13
  SPEAKER_02: So uh maybe simplifying uh a a little bit, you're with artificial intelligence, we're talking about making machines to be like people, right?
• 8:21
  SPEAKER_02: Is that fair?
• 8:22
  SPEAKER_01: Yeah, absolutely.
• 8:23
  SPEAKER_01: Yeah, yeah.
• 8:24
  SPEAKER_01: And and so one of the questions we could ask um is could machines ever be more intelligent than human beings?
• 8:32
  SPEAKER_01: Okay, well let me ask you, what what do we mean by intelligence?
• 8:35
  SPEAKER_01: What does it mean for what does it mean for a machine to be more intelligent or less intelligent than a human?
• 8:40
  SPEAKER_01: What does intelligence mean?
• 8:41
  SPEAKER_01: I I guess the question is could that Siri device, could that ever be human?
• 8:45
  SPEAKER_01: What's what's different about me asking Siri a question and it trying to interpret or you know, extract the meaning from what I ask it?
• 8:52
  SPEAKER_01: What's what's different about Siri than if I were to ask you the same question?
• 8:56
  SPEAKER_01: You know, what what is the temperature outside?
• 8:57
  SPEAKER_01: And and these sorts of questions get at the m the the interesting debates and topics in AI is what does it what does it mean to be sentient?
• 9:06
  SPEAKER_01: What does it mean to be human?
• 9:07
  SPEAKER_01: What does it mean to be intelligent, to be conscious?
• 9:11
  SPEAKER_01: Um and can these systems ever exhibit such things?
• 9:15
  SPEAKER_01: One of the main questions that I have are could we ever create systems that are have such a thing?
• 9:20
  SPEAKER_01: And the and the answer, of course, I I would say is no.
• 9:24
  SPEAKER_02: How do you get involved with with this type of uh study?
• 9:27
  SPEAKER_01: So uh I'm not doing any active research uh in in the AI field right now, but I uh was first introduced to the field in grad school where I earned my PhD in mechanical engineering.
• 9:40
  SPEAKER_01: Uh essentially at that time we were researching uh a protein called von Willeburn Factor, uh which is pivotal in the blood clotting process.
• 9:48
  SPEAKER_01: Um most humans have it.
• 9:50
  SPEAKER_01: If you don't, you have uh you're you're very likely to have some sort of blood clotting disease.
• 9:54
  SPEAKER_01: Um but in short, what we were doing is we were performing uh Brownian dynamic simulations of this protein to try to elucidate its behaviors as it flows throughout the body.
• 10:05
  SPEAKER_01: Uh so we performed these simulations and and got out of these simulations very complex data that was hard to understand and it's it's hard to analyze that data.
• 10:15
  SPEAKER_01: So we used um machine learning algorithms, which is very closely related to artificial intelligence, but we learned some machine learn, use some machine learning algorithms to try to better understand our results and extract patterns that we wouldn't have otherwise been able to see to help us answer the question uh essentially, what does this protein need to look like?
• 10:37
  SPEAKER_01: What does the general macromolecular structure of this protein need to look like in order for it to uh facilitate its uh responsibilities in the blood clotting process?
• 10:48
  SPEAKER_01: Right?
• 10:48
  SPEAKER_01: So that that was the question we were asking, and and we used some artificial intelligence algorithms to help us do that.
• 10:54
  SPEAKER_02: What are some of the more uh fascinating discoveries about in this field that some of our listeners would be well familiar with?
• 11:02
  SPEAKER_01: Popular uh in our culture now is uh discussion about self-driving cars.
• 11:07
  SPEAKER_01: And um one fascinating discovery that uh I've been following uh is is just self-driving cars.
• 11:14
  SPEAKER_01: And in 2014, uh Google made the first uh self-driving car to pass a state driving test, and that was in Nevada, and we've made tremendous uh improvements since then.
• 11:25
  SPEAKER_01: Um I think of companies like Tesla, um just how far that they've come in that technology.
• 11:30
  SPEAKER_01: Uh also I've been filing um are are you familiar with the the it's an ancient Chinese game called Go?
• 11:37
  SPEAKER_01: No, I'm not.
• 11:38
  SPEAKER_01: So this is a it it's ancient, it's thousands of years old.
• 11:41
  SPEAKER_01: It's it's a game called Go.
• 11:43
  SPEAKER_01: It's played on a board with stones.
• 11:45
  SPEAKER_01: Um it's a 19 by 19 board, and um stones are played on the uh the vertices of this these 19 by 19 grid.
• 11:56
  SPEAKER_01: Uh, and it's an incredibly difficult game.
• 11:59
  SPEAKER_01: The the possible move space, so how many moves you could make in this game, there's there's something like 10 to the 137 uh possibilities.
• 12:08
  SPEAKER_01: So it's it's it's far too large of a system uh to be fully known by any any algorithm.
• 12:14
  SPEAKER_01: So the reason I think that this is interesting is because um the late some of the latest advancements in artificial intelligence have been with this game.
• 12:22
  SPEAKER_01: And although those advances have just been in a game, the techniques that uh they have employed to bring mastery uh of these algorithms uh in playing this game are transferable to other technologies.
• 12:36
  SPEAKER_01: So so uh there's actually a documentary out right now, uh I believe it's called Alpha Go, and it's a whole story about the the technology, the artificial intelligence algorithms that have gone into uh this work to create algorithms that can excel at the game of Go.
• 12:53
  SPEAKER_01: So if you want to check that out, that would be Google's Deep Minds, Alpha Go.
• 12:57
  SPEAKER_01: So Alpha Go is the name of the algorithms that that beat the world champion in in the Go game.
• 13:05
  SPEAKER_01: Also in 2019, uh company called OpenAI, which uh created uh a robot hand to solve a Rubik's Cube puzzle with tremendous dexterity.
• 13:15
  SPEAKER_01: And um this is this is a cool topic here because it it aligns with your previous definition of artificial intelligence, where some of these algorithms, you are able to use these algorithms without ever explicitly writing some computer programming.
• 13:31
  SPEAKER_01: So what that looks like is essentially you you have a you have a machine and you you give the machine some information.
• 13:39
  SPEAKER_01: Uh the machine takes that information and tries to make a decision.
• 13:43
  SPEAKER_01: Go left or right, or choose yes or no.
• 13:46
  SPEAKER_01: Um and it it doesn't necessarily have any basis for making that decision, but it will choose one option or the other.
• 13:53
  SPEAKER_01: Uh and based on that option, the the programmer is able to tell the machine whether it did a good job or a bad job.
• 14:00
  SPEAKER_01: And based on this system of receiving information, making a decision, you know, taking some action, and then being told if that decision was good or bad, the machine can eventually learn, just like humans learn, to consistently make the right choice.
• 14:14
  SPEAKER_01: And so this is a branch of artificial intelligence called um reinforcement learning.
• 14:20
  SPEAKER_01: And reinforcement learning is what was used to train this robot hand uh to have such tremendous dexterity in solving a Rubik's cube with one hand.
• 14:30
  SPEAKER_01: Um and the interesting thing about that, too, is that all of that training actually occurred in simulation and not in the real world.
• 14:38
  SPEAKER_01: So they fully developed uh software that would enable uh a robotic hand to manipulate a Rubik's cube and solve it without ever practicing in the real world.
• 14:51
  SPEAKER_01: The first time that they tried it, it it just worked, um, which is pretty incredible to me.
• 14:55
  SPEAKER_02: So it it sounds like from what you've been saying that uh uh machines can be more intelligent than some of us a good percentage of the population.
• 15:06
  SPEAKER_02: For example, uh not all of us uh can drive uh a car right now.
• 15:10
  SPEAKER_02: Either we're too old or we're too young.
• 15:13
  SPEAKER_02: A larger percentage of the population can't solve a Rubik's Cube with two hands, let alone one hand.
• 15:20
  SPEAKER_02: So you're saying that these uh that artificial intelligence is truly very intelligent indeed.
• 15:25
  SPEAKER_02: Well, how do you define intelligence?
• 15:27
  SPEAKER_02: I guess I'd say it is the uh ability to assess a situation in front uh uh of one, uh uh to be able to evaluate it and then to react and respond in an appropriate fashion.
• 15:41
  SPEAKER_01: So this is this is a huge topic in the field and and what what constitutes an intelligent machine.
• 15:47
  SPEAKER_01: Um and actually kind of the benchmark test in the field is what's known as the Turing test, which was established by Alan Turing um in the 50s.
• 15:56
  SPEAKER_01: And actually, Turing proposed that the test would be natural language, um, and that if a machine could understand natural language, that would be a good measure to test if it was truly intelligent or not.
• 16:10
  SPEAKER_01: And so um Turing explains that there's there would be uh actually what is called the imitation game, um, an evaluator, and this evaluator would be a human, and that evaluator would be blind to two other people.
• 16:27
  SPEAKER_01: Uh well, excuse me, one other person would be a human, and another would be a machine.
• 16:32
  SPEAKER_01: And basically the evaluator would be able to communicate with this other person and with the machine only through uh text.
• 16:40
  SPEAKER_01: So perhaps a keyboard and a monitor to communicate with these two other, you know, the person and and the machine.
• 16:46
  SPEAKER_01: And the Turing test is if the evaluator cannot reliably tell the difference between who is the human and who is the machine, then the machine passes the Turing test and is would be considered as intelligent.
• 17:01
  SPEAKER_01: Um and since then, actually, the field has also developed what is called the total Turing test, which kind of extrapolates that idea to also include spoken word, right?
• 17:11
  SPEAKER_01: So if you could have a conversation with the machine and not be able to distinguish who the human is and who who the machine is, then that would be a truly intelligent to scene uh machine.
• 17:22
  SPEAKER_01: Uh and also fairly there's also been the the truly total Turing test, which kind of adds a layer on to that as well.
• 17:29
  SPEAKER_02: These uh machines can be very uh intelligent.
• 17:31
  SPEAKER_02: It seems to me that they can be used for good and it could be used for evil.
• 17:35
  SPEAKER_02: Why don't we talk about uh some of the good right now and interviewing you during the time of uh the coronavirus and uh the uh lockdown we're having in this area?
• 17:43
  SPEAKER_02: Uh can artificial intelligence be used uh to help treat uh human beings that are suffering from this virus right now?
• 17:51
  SPEAKER_01: Yeah, so there's two ways that I kind of think about this question.
• 17:54
  SPEAKER_01: There's one in the in the the sci-fi AI way of thinking of this artificial general intelligence machine that is a nurse robot, right, and is able to take care of us.
• 18:05
  SPEAKER_01: And of course we're not there yet, but what I would like to say about that is that Christians shouldn't be afraid of that idea, right?
• 18:12
  SPEAKER_01: And that all of AI, these are these are tools that we are creating, and we are creating these tools, they could be used for good, and they have good application, and and they have harmful application.
• 18:23
  SPEAKER_01: And to cast that into other words, these tools can be used for human flourishing, or they can be used in ways to that that that hurt human flourishing.
• 18:34
  SPEAKER_01: And and generally speaking, the idea of having some sort of cyborg nurse, although it's in the future, decades or more years out, that is not something that Christians should be fearful of.
• 18:46
  SPEAKER_01: And that is something that I think that we need to be embracing, but be very aware and cautious of the ways that this sort of technology could be misused, right?
• 18:55
  SPEAKER_01: So that's one thing I'll say.
• 18:57
  SPEAKER_01: Uh secondly, I'd say a lot of the there's a lot of modeling, statistical modeling, mathematical modeling, and and and modeling done with these algorithms.
• 19:07
  SPEAKER_01: And the models that have been created through machine learning and artificial intelligence algorithms, those shirt certainly should be trusted and have are very informative in trying to understand the spread of the coronavirus and and modeling of infectious diseases.
• 19:24
  SPEAKER_01: So so uh clearly that is that is a phenomenal way of using some of these tools, right, is to better understand the spread of these diseases.
• 19:33
  SPEAKER_02: You could use it uh then, let's say, to theoretically uh uh create modes of transportation that uh wouldn't wreck one with another, right?
• 19:41
  SPEAKER_02: I mean that would be uh uh a means uh that would be possi positive uh to send uh uh in a representative uh into a a ward that is filled with uh a virus, a highly contagious virus, uh if you had uh uh uh artificial intelligence, uh you could do that, uh correct?
• 19:59
  SPEAKER_01: Absolutely, absolutely.
• 20:01
  SPEAKER_01: And and and I I don't think anyone sees an issue with that, and and and that is a good implementation of of these technologies.
• 20:09
  SPEAKER_01: But I think it's easy to see just a short step away that you could create, just just as though you could have an autonomous caregiver, which is what we're describing, you could also have an autonomous weapon.
• 20:20
  SPEAKER_01: And so I think these this starts to touch on some of those ethical ramifications of what artificial intelligence is.
• 20:25
  SPEAKER_01: And and the idea of a lot of these times with you know with AI, it could just be software, but it's also typically coupled with hardware to create these autonomous systems.
• 20:35
  SPEAKER_01: And what do Christians think about having an autonomous weapon, right?
• 20:39
  SPEAKER_01: I mean, this is is there just war?
• 20:41
  SPEAKER_01: This is a whole theological discussion that we could get into on its own merit, but um this is why Christians need to be involved in this this field, right?
• 20:49
  SPEAKER_01: We can't just we need to be sure that as a society, we're not we're not creating uh a duality where you know there's church and there's choir and there's there's your church community over here and one one party to life and there's science and politics and everything else over here and and those two are are separate.
• 21:07
  SPEAKER_01: We need to make sure that we're not doing that, but but rather that we have Christians working in the field of artificial intelligence and that and that we have people who are designing systems in in such a way that um hopefully they won't be used for such egregious uses, um, such as autonomous weapons and and things of that nature.
• 21:25
  SPEAKER_02: So artificial intelligence uh could lead to a good decision, but then it could also eat lead to a bad decision if we leave um robots, uh machines, uh and the ability to think for themselves.
• 21:39
  SPEAKER_02: Furthermore, if it's programmed by somebody who has morals, it might lead one direction.
• 21:43
  SPEAKER_02: If it's programmed with some b by somebody who doesn't have morals, it could lead to another.
• 21:48
  SPEAKER_02: Can you speak to that at all?
• 21:50
  SPEAKER_02: So ethics in this.
• 21:52
  SPEAKER_01: Yeah, uh all of these systems, right, artificial intelligence systems, are in the end of the day, they are decision-making systems, right?
• 21:59
  SPEAKER_01: They make decisions.
• 22:00
  SPEAKER_01: And take actions.
• 22:01
  SPEAKER_01: So whose decisions are they?
• 22:04
  SPEAKER_01: When a machine makes a decision, who bears the responsibility for the results of that decision?
• 22:09
  SPEAKER_01: And I think that is where the real discussion is had.
• 22:12
  SPEAKER_01: So there's no doubt that we as a society, we of course could create an AI system that's capable of killing.
• 22:19
  SPEAKER_01: We of course could create an AI system that's capable of misleading people, uh, of of being of preying on them, right?
• 22:28
  SPEAKER_01: And but the question is, not if we could create these, but is there ever a permissible scenario where killing using AI would be okay?
• 22:37
  SPEAKER_01: Is there ever a permissible scenario where some sort of you know another act would be okay?
• 22:43
  SPEAKER_01: And I think that that is the ethical debate.
• 22:45
  SPEAKER_01: Imagine fighting wars with fully autonomous systems against people who plausibly would not be able to do that, right?
• 22:53
  SPEAKER_01: Sure, it it it may minimize the loss of life, but is that an honorable way to fight war?
• 22:59
  SPEAKER_01: Is that if if fighting war is okay, is that how Christ would have us fight war?
• 23:04
  SPEAKER_01: Where the one group has no risk and the other group bears much risk, right?
• 23:10
  SPEAKER_01: Where where somebody flying a drone can't be hurt, but but the person on the other end can.
• 23:17
  SPEAKER_01: Some of the other ethical issues related to AI are privacy and surveillance, right?
• 23:22
  SPEAKER_01: So these models, generally speaking, are trained using enormous data sets.
• 23:27
  SPEAKER_01: And these data sets often include information about you and I, right?
• 23:32
  SPEAKER_01: Things that are taken from our Facebook accounts or Amazon, all of all of these different areas, right?
• 23:37
  SPEAKER_01: So imagine with me the scenario where there's a model, an AI, an AI algorithm that's trained using a huge data set that incorporates our data.
• 23:48
  SPEAKER_01: So imagine a malicious user, um, and and this has actually been shown that what a malicious user can do is take that model that's been trained using an enormous data set and query that model or give it some input parameters in such a way that it's able to find out identifying characteristics about the training data set.
• 24:10
  SPEAKER_01: What that means is there could be companies like Facebook and Google that are using our data to train models and then making those models publicly available.
• 24:21
  SPEAKER_01: And the problem is that some bad actors could take those models and get information about UNI.
• 24:27
  SPEAKER_01: And so what if what if the model that was used, hypothetically, contained our names, our social security numbers, and maybe our credit card numbers, right?
• 24:36
  SPEAKER_01: So so it is possible to, from the back end, work out identifying information about UNI, right?
• 24:44
  SPEAKER_01: So that this is why privacy and surveillance in these systems is such a big discussion.
• 24:48
  SPEAKER_01: And I mean, you can also look at the mass surveillance that China has implemented of its people.
• 24:54
  SPEAKER_01: Um and in fact, just last month, uh, a Chinese company said that it has developed facial recognition technology that can identify people uh even when they're wearing a mask, right, during coronavirus.
• 25:06
  SPEAKER_01: That that's astounding, right?
• 25:08
  SPEAKER_01: That that there's uh a facial recognition software that can only see the a third of your face and still identify who that person is.
• 25:15
  SPEAKER_01: That's that's incredible.
• 25:17
  SPEAKER_02: So the uh becomes then important for those with a moral backbone not to sit on the sideline for this.
• 25:25
  SPEAKER_01: Right, yes, yeah.
• 25:26
  SPEAKER_01: And that's that's why we need to get we need to engage in the conversation now before these problems are already full blown in society.
• 25:33
  SPEAKER_02: And what what does what does that mean to engage in the conversation and and where would people with a moral backbone do so?
• 25:40
  SPEAKER_01: That is why we need people to who are going to get higher educations, and that's why they need to come to Eastern to do it.
• 25:46
  SPEAKER_01: You know what I'm saying?
• 25:50
  SPEAKER_01: No, but seriously, that's kind of that's kind of we need Christians that are gonna go get higher educations, and and we need Christians that are gonna go be in this field and design these technologies.
• 26:00
  SPEAKER_01: Um, you know, there are there are end users of these systems and there are developers of these systems.
• 26:05
  SPEAKER_01: And we need to make sure that we're we're getting godly people developing these systems.
• 26:11
  SPEAKER_02: And for those uh who uh don't have the opportunity to get an education in this, what do you feel that um let's say the church's role ought to be in this?
• 26:18
  SPEAKER_02: Are you saying uh suggesting not only support, but uh are you suggesting other means?
• 26:23
  SPEAKER_01: Yeah, uh is just to engage at least engage in conversation and not allow this to be a topic that is something other than church, something outside of church.
• 26:32
  SPEAKER_01: But rather that that the church is relevant, we are here today, and that we are going to face the issues of today head on and talk about them and and not let them be something other.
• 26:42
  SPEAKER_01: That is what I mean to say.
• 26:44
  SPEAKER_02: Well, thank you very much, uh Mike, for being our uh guest today, and thank you to our listeners uh to listening to uh Centrally Speaking, a broadcast ministry from the Central Schwankfelder Church here in Lansdale, Pennsylvania.
• 26:55
  SPEAKER_02: Our website is www.cscfamily.org.
• 27:00
  SPEAKER_02: If you're in the area, we invite you to our Sunday morning services, a traditional service at 9 o'clock AM with organ and various choirs.
• 27:07
  SPEAKER_02: Our praise team leads a contemporary service at 10 35.
• 27:10
  SPEAKER_02: Sunday school, where we talk about issues uh such as this, um uh happens at 9 and 10 30 for adults, and there's a youth Sunday school at 9 o'clock uh a.
• 27:19
  SPEAKER_02: We warmly welcome you to join us.