"Into the AI Flood" by Benjamin Harris

The one may be overpowered by another, two can withstand him, and a threefold cord is not quickly broken. This is Strength to Strength, and it's good to be with you, the ones on the call, and also our listeners, um, here again today.

And we're anticipating looking into a talk on artificial intelligence, and we'll get into that, some of the details of that here right shortly. Um, so Strength to Strength, it's our vision, it's our heart to, for us and for our listeners, for those who are participating, to grow in their faith, to go from strength to strength, as the psalmist says so beautifully in Psalm 84. And our vision, our our the destination that we want to head towards here is to inspire flourishing communities of faithful Jesus followers around the world. And a verse this morning that's on my mind is out of Matthew 6:33. Seek first the kingdom of God and His righteousness, and all these things will be added to you.

Seek first, pursue it, seek it, um, search it out, mine it out. Um, and there's just so much in that verse. And, you know, if we could just, um, it's our heart that as as we meet here, as we hear talks that challenge us, as we connect here, is that we could just in a in a better way, in a deeper way, um, imagine what God's kingdom is, understand it, um, see it in scripture, understand the laws of the kingdom, uh, of God. Um, but I think too, like I mentioned earlier, just is is imagining it. What what would it look like, um, to be in truly in God's kingdom, the new heaven and the new earth, whether it be the Garden of Eden, um, or, um, or the the the city, this uh new Jerusalem, this this heavenly city, um, or the Garden City as we see it there in Revelations.

What would it be like to be in that? And then to pursue that as as followers of Jesus, to pursue that, to bring that to earth now, um, in our in our communities and with others, not alone, but with with our church community. Um, that's that's the heart of Strength to Strength.

So yeah, shortly I'll introduce, um, Ben Benjamin Harris, or Ben, uh, as he said we can call him, um, and and and and his talk. Before that, we have a couple announcements, and then of course we'll have prayer after this as well. So the first announcement is at Strength to Strength, we've decided to, for the summer month, June, July, and August, to just step back from our normal cadence of every two-week, um, talk here and go to once a month. Um, as some of you know, we have a a sisters, a Strength to Strength sister side, and they do the first Saturday of the month, 12 talks a year. Um, and of course we do every two weeks.

But but we plan to kind of match that cadence, um, here for June, July, and August, and just slow down a little bit, um, on this platform, especially on our talks. Obviously, we have a whole book side that'll keep on moving. We have some books, um, in the works, and and I'll talk about it here right shortly on my next announcement. But yeah, so that'll that'll keep moving over there. Um, but here, we'll just slow down. Um, a lot of us, um, a lot of our listeners, of course, us at the admin, we have, um, it's summertime, right? And so here, especially here in in in North America, um, this this part of the world, and there's a lot of a lot of summer activities.

And and, uh, sometimes a, you know, a regular 6 a.m. Saturday morning call can be a little challenging to get to. Uh, and of course, for the admin team here, just kind of keeping up with that, uh, we felt like, you know what, it's actually will be good just to slow down, step back, reflect, uh, and then prepare, uh, prepare for, um, going back to a regular two weeks, Lord willing, uh, in September. So, um, yeah, so next Saturday, uh, will be June 1st, I believe. Uh, and so that'll be our, we'll have another another talk next Saturday, and then we'll break for about a month and beginning of July, have our next talk. So yeah, we wanted to make sure that that message gets out here, and of course, we'll communicate in other ways as well.

So, all right, and then for the second announcement here, announcement here, um, this will be for more of the S2S books or Strength to Strength books side, um, is we have a new, uh, a new series, uh, teaching series that we that we're going to be bringing to our our bookstore in the future. It's still a couple months out. Um, Glenn is going to kind of give more of those details, but it is a, um, it is a product called Salt and Light. And Glenn, if you want to just pop up that first slide there, um, about this, about this product, um, Salt and Light, um, material, um, it's in collaboration with Christian Aid Ministries and Strength to Strength.

We're really excited about this. Um, it's a Bible-based financial teaching course. And as many of you know, would know of Salt, uh, which is a project within CAM. Um, and Salt stands for a Saving, um, Accountability Lending Teaching Program. And it's an initiative there where many of us would know them as as the ones who are doing savings groups in other parts of the world, in developing countries. And so for the last 10 years, they have they've been building this, um, uh, this initiative, working with, uh, Anabaptist, um, workers who are working with local, um, uh, local facilitators who who are taking this teaching, this biblical teaching into poor communities with with much physical poverty and helping these people understand, get a biblical worldview and how they can steward their resources to rise out of poverty.

And so when I think of, uh, of Salt, um, I think of that, but Salt has a number of other areas that they're working on. Um, and and and this, uh, Salt, um, or Salt and Light curriculum is one of those. And they've been working on this curriculum for the last four to five years, um, holding, um, you know, uh, seminars, getting out the materials, um, here in North America where there's churches and people that are interested in reaching their communities with this. And so anyhow, so this this is a a a curriculum that has been in development now for four or five years. Like I mentioned, they've put tons of work, tons of research into it.

And, uh, they reached out to us here at Strength to Strength, wanting to know if we could actually publish it and also ship it and then also help kind of, um, bring together educational talks around it, um, so that people can understand how to use it. Uh, and so we're really honored and excited to collaborate with Salt, um, with this curriculum. And so just a little bit, um, of of of, uh, how to help us understand why you would use it. Um, I'm just going to read a couple of bullet bullet points here. As you consider how God is calling your congregation to demonstrate God's kingdom locally, we invite you to consider the following questions.

Are you struggling with affluence and its tendency to draw us away from God? Do you have a burden for youth who have so many options where they can invest their lives and resources? Are there young families in your congregation who could benefit from stewardship teaching or personal finances?

Are you an employer who longs to provide sustainable help to your workers and who struggle for and who struggle from paycheck to paycheck? Are you assisting those who struggle with finances or broken relationships? Do you desire to provide meaningful, sustainable help? And I think sustainable help here is really at the heart of Strength to Strength, providing more than just a handout, right, but rather a hand up. And so if you answer yes to any of these questions, check out the Salt and Light courses to help you consider God's truth and share them with others with others. Seek to become a faithful steward. And so, um, Glenn, Glenn's going to give us a little more details kind of on this, uh, and then we'll jump into our talk here.

So, Glenn, uh, go ahead. Yeah, thank you, Brian, for what you shared there already. So, um, like Brian said, we're in, um, you know, in preparation for this. It's not available yet. It's still, uh, being finalized, and, um, it has been in development for a number of years, but the development phase is coming to an end. And, uh, we're working towards, um, making this available, uh, towards the end of this year, possibly November or so.

So, um, it is designed for a classroom setting. This is not something for you to purchase for a self-help program, but it's designed that there's a teacher and a class. And so that's, um, you know, kind of the framework to to keep in mind with this.

Uh, there's four tracks, and there's going to be much more detail on these, but track A is called Chronic Need to Assist Those Living in Chronic Financial Difficulty. That's actually where this had started from initially. And then the, uh, during the development, they decided, well, we're going to go broader than this. And so it has, um, developed into four tracks. As you can see on the screen, track A is Chronic Need, B is for Local Community, C is for Local Church, and D is for Young Adult. And, uh, each one of these has a participant guide, which is like, if you will, the the student, and also a facilitator guide, which is for the teacher.

And the facilitator guide will include the text from the participant guide plus additional helpful resources. So that's, uh, that's the main thing there. Uh, here's just, um, a testimonial from somebody who has been a part of the pilot program and, uh, took the course, the Salt and Light course, taught me how to have a healthy relationship with money. And, um, it's the the course is built around financial training, but it's an excellent, uh, platform to use then to, um, you know, to work on discipleship. And, um, so yeah, we're we're excited about this. There's going to be more information upcoming in the coming months and around the end of the year, maybe November. Uh, we're looking for general availability.

So we will, uh, keep you all in the loop on this. Yeah, thank you, Glenn. Um, and if you have any questions, you can hop on, uh, strengthstrength.org and contact us there or email us at contact@strengthstrength.org as well. So thanks, Glenn. All right, now for what we're here for this morning, um, on this topic of artificial intelligence, or as Ben has titled this talk, into the AI flood. And so many of us would think of 2023 as kind of being the year, a year ago, of where this really kind of hit, um, pop culture, this, uh, this artificial intelligence. Many of us are like, wow, what is this? And, but Ben, uh, I'm sure, um, has been quite aware of this for a long time.

This isn't just a new thing. It's been around for decades, as I understand, but I'm not going to act like I understand it, so, because I don't. Um, so Ben comes to us. Uh, he's a PhD professor, uh, currently associate professor of business at Sattler College in Boston, Massachusetts.

Prior to Sattler, he was on the faculty of Boston University, before which he spent more than a decade working at a large technology company in the area of AI or artificial intelligence and machine learning. Uh, Ben is married. He has five children. Ben, it's good to have you with us today.

Thank you for great to be with you all. Thank you. Thanks for coming on here this morning. And this is how Ben describes the his talk that he's going to bring to us this morning. Artificial intelligence is ushering in a new era of questions, both in the workplace and in our lives.

We briefly work to separate the, we'll we'll briefly work to separate the reality from the hype. Finally, we'll touch briefly on how AI can be a double-edged sword in our lives and things we should consider in stewarding our work and our lives well.

And so, Ben, I'm anticipating this this talk. We've heard these apocalyptic type, um, uh, discussions, professions, uh, prophecies about what this is going to be. We've also heard these utopian things about what AI is. And so, uh, I'm looking forward to hearing you share. And then, of course, at the end, we'll have some questions for you. Um, so we'll have an open Q&A. And so be ready for your questions for for Ben then. So it's all your, actually, you know what, let's just pause for prayer first. Let's pray, Father, we come to you in the name of your Son, Jesus, this morning. Thank you for your mercy this morning. Thank you for, um, just this peace that we can have in our hearts, knowing that you are in control.

Um, Father, we live in an incredible world. Um, we look at creation and it just proclaims your majesty. We've seen some such beautiful proclamations of that recently with, with, um, from the solar eclipse, um, to the solar storms, created this incredible light display in the sky at night. Father, we've been, we've been so, um, we we just see these things and we realize what a what an incredible God you are. And Father, also we look at humanity, all 8 billion people on this planet, and we see, um, we see so much darkness, but we also see light. We see people who are striving to be faithful disciples of yours. And Father, we we ask, Lord, that this talk this morning could enable us, um, equip us to be better be your disciples.

Father, we know that the evil one is, uh, that in the principalities and powers of this world are conspiring together to, to, um, ruin this world, to, to bring it to, to, to, um, desolation. And Father, uh, and so Father, we know that, uh, with, with technology, of course, just a piece of it's, it's a tool in the toolbox to that we use to destroy us. And so Father, I pray that we will be as wise as serpents and as harmless as doves. So bless Ben as he brings this talk here this morning. And we ask this in Jesus' name. Amen.

All right. All yours. All right. Well, good morning. Good to be with everybody. Glad to glad to be here with you. I join you from the East Coast of the US, Boston area. And, um, yeah, so I get this opportunity really to just spend a few minutes walking through, um, the the truth and the hype that surrounds this artificial intelligence idea.

Um, and our, you know, our goal, I think, is just to think biblically about it, right? There's this, we're not going to be able to avoid it forever. Uh, it seems to invade more and more of, um, the our daily interactions with each other, both on, um, I know even on Zoom meetings, right? We spend time on Zoom meetings, and there's now AI transcription enabled, uh, in many of these technologies, which, hey, we might think is really helpful. Like if I I could read the transcript, that would be fine.

Um, but the question is, what do what do we do and sort of what is the background behind all this? Um, so I'll give you a quick overview here. Um, there's going to be a few a few points we're going to walk through. We're going to look at at a little bit of the history of AI, and you're going to see that this is this is not new. Um, there's we have sort of a new wrapper around all of this, but the the idea of machine intelligence has actually been around for a long, long time. Um, what made it possible now, though, is a few a few accelerators, right? They're like, uh, humans' ability to do certain things had to reach a certain level for AI to become as powerful as it is.

And so, um, so we'll talk about those. We'll look at the current revolution that, uh, you know, what is going on with artificial intelligence in the world today. Um, talk a little bit about worldview, right? This changes how we see the world, how we should instruct our children, how we should encourage one another.

Uh, and I'll I'll just end with a few thoughts of, you know, what do we do about this, right? Is this something to, you know, again, Glenn or Bryant, as you mentioned, uh, do we sort of move in the apocalyptic direction? Should we think of that way? Is it a utopian direction? Uh, I'll I'll give you the, uh, the end of the story first. I think it's neither, right? But it's, um, it's something different. What do we do? How do we steward our time and our work well here?

Um, now the the challenge before I go into the history, actually, I'll back up just one moment. Um, the the trick with AI is is when I was writing these slides, I was struggling because I realized that every day everything I wrote on here seemed to be out of date. That was the the challenge is the new, you know, some AI company would come up with some new development and and what was previously cutting edge is no longer cutting edge. And I said, well, okay, how do we, you know, how do we present something that is real and current? And so I had to back up my lens a little bit and say, what is generally true of this entire revolution?

Um, instead of focusing on individual technologies, individual companies, um, kind of what do we what do we do about this? So, um, the history here, um, you can look way back into antiquity and see this. There was a bit of fascination, um, with something that was, it didn't seem to have the traditional definition of life, but was still alive, right? You go all the way back to, uh, Talos in in the Greek, um, mythological pantheon. And he was a bronze figure who would protect the island of Crete and hurl boulders at ships. Almost a, you know, if you want to think about it, a, uh, a very early antivirus kind of thing. But there was this concept of, um, you know, hey, he's, it's not alive, but it's this bronze statue.

Uh, and in in the mythology, um, Talos was defeated by Jason the Argonaut, sort of if you read the, the, the, um, old mythological tomes there. Um, here they pulled the plug from his foot and he was left inanimate was the the whole idea here. Um, but that wasn't, that story was not, um, in isolation.

Uh, you get similar stories from the Middle East, the Mediterranean region, North America from antiquity to today. Um, there were there were many, many of these stories that described what you might now see as just a curiosity with how, how does something seem to display, um, I guess, forms of life without actually having, um, metabolism, without actually having what we would call life that God has given to us. So this has been around.

Um, now the first, the first forays into this had to do with, uh, the development of the computer, right? We began to, um, take our, the synapses that, that, um, allow our brain to operate, translate them into a board effectively. Um, now this is a picture, the photo I have on the slide here is of a computer called ENIAC. And it was one of the first machines. It was made at, uh, the Princeton University or, um, Institute for Advanced Study at Princeton in the 1930s. So this is preceding World War II. Um, so, you know, if you if you want to think about the scale of time that we're talking about here, all the way back to Greek antiquity, about 2000 years ago, uh, and our ability to to digitally simulate anything even remotely resembling life was less than a century ago, right?

So, so all the the technology that we're talking about is very, very new, right? And as in the in the scale of God's redemptive history plan here. Um, so ENIAC is one of the first. Now, to give you a sense of scale, all that you see in this room, because computers used to be about the rooms, about room sized, all that you see in this room, uh, is now a fraction of the processing power that is on one chip in your phone, right? If you have a cell phone or a computer. Um, so our, you know, we've really, the thing that's changed between the 1930s and today is the size and the speed of processors.

The actual technology has not changed that much. Uh, things just get faster and they get smaller. So computers really began to push this thing forward. Now, what happened, uh, just after this, this sort of era in technological history, uh, is mathematicians began to to think like, ooh, now I have a, um, a device that will, that will take instructions that I give it, and it will execute them much faster or much more consistently than I can, uh, as a person. What can I do with it? And so, um, cognitive scientists and computer scientists got together, and they began to map out what they believed the the brain circuitry to look like. They said, okay, we we think the brain is constructed in a certain way.

Uh, and they began to create mathematical models that they they called early neural networks. They were, um, we now refer to them as ANNs, but artificial neural networks. Um, and you can you can hear in the in the midst of that title, the the the phrase neural, right? So we're talking about stuff in the mind. And neural networks are, um, they're a they can be best described as just a learning architecture, right? They have a single task that they've been trained to do, and you're going to put, you know, thousands or hundreds of thousands or millions of bits of data into a network, and it's going to learn what's right, what's wrong. And then at the end of the day, you're going to just use it to make a prediction.

You're going to use it to make an inference about what reality is in front of you. So that mathematics was developed way back in the 1930s and 1940s, right? This idea of a neural network. Now, the problem was they didn't have a fast enough computer to be very effective with it. So often, so the the they would write down this mathematical model on paper, and they'd have no place to put it. They'd have no language to program it. They didn't have enough computing hardware to do anything with it. Um, and that stayed largely the case until we got something called a GPU. Um, even CPUs, um, you know, got much faster. The, and actually, let me back up for a moment.

I'll define the difference here. So in most of the technology you use, so your laptop computer, your desktop, uh, your phone even, uh, these typically have what's called a, um, a central processing unit or a CPU. And it's a classic computer chip, right? We have these in all of our devices. Now, what they do is they do what we call floating point mathematics, right? They can multiply, they can do operations on two numbers, right? They add them, subtract them, multiply, divide. And they do, they do all these operations in a line. And they do them really fast, right? So we we almost recognize that it's almost this instantaneous processing.

Um, but you can take your desktop computer and you can push its mathematical capability to the limit and see, okay, it actually does slow down at a certain point, right? There's a there's a limit to how quickly we can do math on a CPU.

Um, so early neural networks were, uh, they worked fine. They worked okay on CPUs, but there was this linear limit they couldn't overcome. They said that they can only go as fast as the chip can do work. Um, so what what a, you know, what a classic engineer would do is say, okay, we're we're limited in speed by this chip. But what if we had two chips? So they put them next to each other and they and they began to speed up, you know, roughly twice as fast. What about four chips? And they went four times as fast. Um, now, again, that's that's good and helpful, but it's this linear growth. They couldn't do anything much faster than that.

Um, now along comes this idea of a GPU. Now, a GPU is a is a a graphics processing unit. Now, these were originally, um, video cards in computers. So we talk about, you know, being able to render video. Um, and the difference here is instead of doing lots of math really fast in a line, uh, a GPU does mathematics, uh, in a matrix. It actually takes, you know, essentially a two-dimensional, uh, array of numbers and it does mathematics on those. And it does them incredibly quickly. Now, a G, the you can't really interchange the two on a computer. You can't, unfortunately, that would be great if you could. You couldn't take your desktop computer and plug in a GPU and do the same thing you're you were doing.

Um, GPUs are fairly slow at linear mathematics, but they're incredibly fast at matrix math. Now, most of the machine learning and AI models we use these days, uh, involve matrices, right? Um, can anyone remember ever having to, um, if you go onto a website and it asks you to, you know, verify you're not a robot, right? You've probably been through these CAPTCHAs, these these processes. Um, have you seen the, have you been through the CAPTCHA that has, you know, click on all the pictures of a stoplight or a bus? Can you remember those? Absolutely. Yeah. So the the reason for that question is, well, what's that what's that doing? Does that really prove my humanity?

I can click on pictures of a bus. It kind of does that. But one of the other things that it does is it allows the these CAPTCHA companies, Amazon's the most notorious one that does this. It allows these companies to have new sets of labeled data. So when you're doing that CAPTCHA, what you're actually doing is you're you're taking a small photograph on a website and you're telling Amazon there's a bus in that picture. And so that picture then feeds back into Amazon's image recognition models that are GPU operated. Um, and it allows it to learn a little bit more about how to discern objects in a photo. Um, that's why they get so good at these things is they have millions of users who just for a few seconds are helping them label all the data that they need to use these GPUs, um, to their full capacity.

So a GPUs, the actual use of these for machine learning, this is very, very recent. This is only the last couple of years. Um, and most of these, um, yeah, actually the, and tell you the sort of the the interesting part about this is this changed the technological market space. It used to be that, uh, Intel and AMD were the these companies that sort of ran technological hardware, right? Intel exploded. They had their sort of glory days as a computer chip maker. Um, but what became clear, this was probably 2022, that whoever had computing power, right? Whoever, so there's this fixed sort of, I want to don't want to say a fixed pie, but there's this, um, whoever controls the computing power is the one who gets to dictate the future of artificial intelligence, right?

Because you're, as models are growing, we have more and more data. Um, whoever the whoever the person is who controls all that hardware, they get to be the boss of sorts. So what happened was, um, we had these cloud providers. And when I say cloud, uh, I won't assume a technical background from us. So a cloud, all that a cloud is, um, in the computing sense is a a large computer that sits somewhere else that you get to, uh, I can log into and I can do very complex or memory intensive kind of program work. Um, but it doesn't have to sit in my home, right? The nice part, Amazon or Google will manage it.

And I can just use it for 10 minutes and I can rent it by the minute. Um, so that's a cloud, that is a cloud machine. And these big companies all position themselves in the AI space. So now if you have a, um, you know, an AI task you want to do, you download this really big model, you have this task you want to to complete, you can send it to a Microsoft machine using their cloud services. Uh, you can, it'll probably cost you a couple of dollars and it'll return the result in about 10 minutes. Well, depending on the job you're doing. But, um, but that all is is very much available and easy now.

Um, but that wasn't the thing that changed the market fundamentally. The thing that changed the market, um, was the advent of the of the new generation of GPUs. The real growth here came in the hardware is that as as smart and large as Microsoft, Google and Amazon, and there's other companies, um, smart as they are, they didn't have the hardware to run this kind of thing. Um, the company that put all their eggs in this basket that said AI is going to be the whole thing was NVIDIA. So now NVIDIA is worth more than $2 trillion in as a market capitalization. Uh, this is their stock price over the last couple of years, right? So you can you can sort of see how, um, and the scale on the bottom here isn't very good, but it's roughly from, um, my little photos here, 2010 to 2024.

So this first little blip in sort of in the middle of the graph is around 2020. Um, but as G, as the hardware becomes more and more increasingly important, uh, NVIDIA's stock price is just skyrocketing. So, uh, it's now one of the largest companies on the planet. Um, they are, if you had to categorize them, I would say they, NVIDIA is AI optimistic to a fault, right? NVIDIA thinks that the, the, uh, society is is going to increasingly rely on AI. And even the CEO of NVIDIA said that, um, his company with AI technology is going to put the computer programming industry out of business. Um, now, I think, in my opinion, I think that's overblown.

I think computer programmers are still going to be very necessary because AI makes all kinds of mistakes. Um, but this is kind of the, um, I want to back, if I'm going to think back a couple of slides here.

The thing that had to come to fruition before AI could be real, um, was the hardware to drive these things. It was to take these mathematical models that that engineers and scientists have developed and actually make them usable. They, so computers no longer existed just in these national laboratories. Um, you now have more, um, you have more processing power on your cell phone than the computers that put men on the moon did. Um, so once the once the computers became powerful enough, AI became ubiquitous enough, and we find ourselves in this world where AI comes in everything.

So, uh, a couple of definitions, these are, and I I threw this in here, um, just to separate some of the the the words you might hear, and they're a little slightly different. Um, so artificial intelligence, we've been talking about this a bit. All that this is, is simulated intelligence, right? It's decision making. Um, what they what AI researchers refer to is there's the there's kind of two steps in the AI world that one is one is called training, right? So they take, um, an AI system will take in all kinds of data and it'll, uh, and an AI engineer or researcher will have to, they'll direct a computer program at the data and say, this is what I want you to look for and learn about all the things, these patterns that you see in the data.

Computer will go and diligently do the work and it'll learn stuff. The second part, and this is what we experience in AI normally is called inference. And all that we typically experience in the world of AI is a statistical engine that is trying to infer the answer to some question for us, right? It's there's no, uh, there's no real fixed right or wrong in an AI system. Um, all that we really experience is this, um, is this kind of guesswork, you know, if you go to ChatGPT or some other system like that and you ask it a question, uh, it'll give you what thinks is the statistically most likely correct answer, right? And sometimes that's right, right?

They they get it exactly correct some of the time. Um, but if you notice at the bottom of in the ChatGPT window, if you ask it a question, uh, it'll give you an answer and you can say regenerate, right? If you ask it a, it's an interesting experiment to do. You go to ChatGPT and it gives you an answer.

Um, ask it a hard question. You know, you ask it a theologically dicey question. Uh, it'll give you what it thinks is the correct answer. And if you click regenerate, it may give you the opposite answer. Um, so it doesn't mean it's right or wrong necessarily. It just points to the nature of the system that it is a, um, computer system that's been trained on a ton of data and it's inferring or trying to guess at the the right answer for you.

Um, so that's what that's the a good working definition of AI. Um, machine learning, right, is a is kind of a companion set of tools and techniques. The the internal algorithms for AI that it does to infer stuff are machine learning techniques. So you can almost think of machine learning as a as a bubble inside of the bigger AI bubble. Um, and an AI researcher, someone who works in AI specifically, is largely going to be finding new data to train an AI model on. A machine learning person is going to do the the lower level tools and techniques and algorithms kind of work. They're going to say, okay, I need this, I need my computer to learn faster.

How do I do that? They'll they'll change the tools and techniques. Now, the thing, the the apocalyptic nature here is the third definition I have on this slide. The thing that most people, um, are scared of and, you know, and back in my in my younger days, I watched movies I probably shouldn't have. And one of the movies I watched was the Terminator series. Um, and so in the in the Terminator series, there's this this uh malevolent AGI in this series of movies, this artificial generalized intelligence, essentially a computer system that has established sentience, right? It has its own, it can think on its own, it has its own sense of identity.

We are not there yet, right? As much as you might see that, you know, little hypes or headlines that um that we've reached the point of AGI, we are not there yet. I don't think you'll find a credible AI researcher who would claim that we're there.

And the and I think there's a good reason for that. I am you have different opinions on the AGI question.

My own opinion is, again, a sample size of one, it's just one person, is that we are at least a generation away from even considering being close to this. There's some there's some really foundational issues that we haven't solved yet to get ourselves toward AGI. So do you have to worry about the the machines taking over? I don't think so, right? There's where the AI can do many things and it can do, it can seem intelligent. But if you ask it some some pointed questions, you'll realize, oh, this is not as intelligent as I thought.

So again, I want to pull us back from the apocalypse ledge. We're not there. We don't have to worry about AGI yet.

Now, the question is in artificial intelligence is where does where does everything show up here? How do you where do you experience AI? And whether you know it or not, you experience this every day. If you send a text message to someone, right, and it it tries to guess what the next word of the text is, that is an that is a small AI algorithm.

If you want to if you want to think about what ChatGPT truly is, ChatGPT is a very fancy, powerful autocorrect system or autofill, auto text. It's the it's a it's predicting what's the sensible next word in a series of words. That's all that it does.

Now, it does it at a world-class level, but it's there's there's nothing more magical than that behind that. So we see it in texting. We see artificial intelligence in predictive analytics, right? If you're more and more companies, um, you know, publicly traded companies, when they have to give their quarterly reviews to stockholders or to Wall Street, they're they're developing these forecasts like, how are we going to do? How are we going to how's our next three months of business going to go? Um, more and more artificial intelligence is filling in the details when companies are doing, um, they're forecasting their predictive analytics.

Another, this is sort of one of the classical examples is spam filtering. Ever wonder how how your your Gmail account or your Yahoo Mail account tells what is a spam message from what is not a spam message? Um, that's an art of that's a machine learning technique, um, called classification. And, um, I don't know if anyone can remember back to the early, early email days how good, uh, spam filtering was. I can tell you it was not good, right? You'd get all these, I remember the the frequency of getting emails from foreign royalty who had tons of money to offer me was very, very high. Um, fortunately, that has changed, right? Spam filtering is getting a lot better.

Um, also shows up in something like machine translation. Um, I love Google Translate, right? I think it's kind of it's kind of fun to take a, you know, what would that phrase be in the Dutch language? Um, now, that whole technology is fairly new, right? To do almost near instantaneous translation from language to language. Uh, it's AI enabled. That's a natural language processing technique.

Um, another one we interact with is recommendation systems. Um, if you ever watch something on Netflix, right? So you watch a documentary or a movie or something, um, you'll notice what comes at the end of the movie. It'll it'll say, because you liked this movie, you probably will like this one, this one, and this one. It'll give you this sort of series of recommendations. And the way it's deciding that is it's it's collecting data on your behavior, right? Um, and you can sort of think how this might work, right? If you start a movie and you maybe watch two to five minutes of the movie and then you switch to something else, Netflix automatically knows, ooh, this might have been a bad recommendation for this user.

We might want to change to something different. And the and the actual recommendation bar will change. But if you watch it all the way through, you'll say, ooh, that they actually seem to enjoy that, right? We'll recommend some stuff like it, sort of move them in the same direction. Um, and the the goofy one that a friend of mine works at Netflix, uh, told me about is that if someone I'll have I'll ask a question here is if if you're Netflix and you see someone watch an entire two-hour movie and then they watch all the way through the credits to the end, what does a computer model would you assume has now happened?

What do you think?

Right? Does anyone watch all the credits? He probably thinks that you stepped away from your screen. Yeah, you either stepped away from your screen. What what's if you're a if you're a tired parent, what's also a a possible answer?

You fell asleep. Yeah. Yeah, exactly. Yep. So so Netflix has a part of their algorithm is if you watch the whole movie and you watch all the credits, they think you might have fallen asleep. And so they're like, well, okay, maybe we won't recommend something quite as boring as this again. So all these recommendation systems that sit in the back end are AI enabled. They have machine learning that runs, uh, in them. But they are but again, the the foundational idea of training on data and inference hasn't changed. It's just a different application of it. Um, chatbots, right? We see these are um these are showing up a lot. Um, does anyone uh try to think what's the one of the original chatbots was Siri on the iPhone, um, which worked through it had its you know, its faults and foibles, but did some helpful things.

Um, now now these are becoming increasingly available, right? If you go to um like I'm a Bank of America customer, I have a Bank of America account. And if you log into the app on your phone, you are immediately faced with a with a um a digital chatbot named Erica. Right? How can I hi, my name's Erica. How can I help you? And you can ask it a question and it might be able to direct you to the right resource.

Now, do chatbots work perfectly well?

No, they of course not. They they can't do everything, right? You know, it's you can you can imagine these sort of um dramatic moments of, you know, uh relationship ends, boyfriend and girlfriend break up and you log into your your banking app for some reason and Erica says, how can I help? And you type Erica a question of, what do I do? I just broke up with my boyfriend or girlfriend.

Erica is not trained on relationship advice. Erica is trained on Bank of America's website. So a chatbot can really only do as much as it's as it's learned to do. But now these show up, you know, every many companies that are digitally enabled have these things built into their internal systems, right? You can um you can talk with an internal chatbot. Um Other places this shows up are face recognition and image labeling. Um You know, if you um fortunately, the IRS did away with it, but there was one year where if you wanted to do your taxes in the United States, the IRS had a system that they would turn the the the computer camera on your um on your computer on and take photos of your face and then do facial recognition in front of you.

Um Now, privacy advocates got a hold of that idea and they're like, you need to stop. And the IRS said, okay, okay, fine, we'll stop. So they don't do that anymore. Um But that's AI and machine learning enabled technology.

All right. Current cutting edge here. The thing that you see most often these days is what's called a large language model. Um Now, these are trained models. Uh They have a unique architecture, but it's um it's really a model that is just much larger in scale and in volume than the stuff we've seen before.

Uh These began to change the game because of their breadth. They basically took a chatbot and they trained it on almost every conceivable subject, right? That's all that these large language models are. Um ChatGPT is the one we probably know best, but there are others. There's a Google presented one called Google Gemini. Um There's a French company called Mistral that puts out open source models.

Now, what they do, and this is what became this is what has driven things forward. All this stuff that you saw before, like all the tools and techniques you had with AI before, now you find them in a single place, right? You can ask ChatGPT to analyze a picture. You can ask it to give you a good argument. You can ask it to I'm not saying do this. I actually, just for an experiment, I asked ChatGPT to write me a sermon.

I said, okay, if I had to preach next Sunday, write me a sermon. What would you what would you say? Here's the topic. Here's the verses I have to preach on. And it did a very uh clinical academic job of sort of expositing things.