Loading summary
Transcript19 lines
- [00:00]
A
Hello, everybody. Welcome back to the Sky Pod, brought to you by Holy Post Media. I am Sky Jutani. Okay, let's jump into it. I like to stay informed. I try to keep up on what's happening in the world. But despite my best efforts, I have to confess that I feel sometimes totally perplexed by AI. The complexity of this issue, the hyperbolic rhetoric surrounding it, it's made it almost incomprehensible to me. I do take some comfort in knowing that I'm not alone. I think that's probably how most of you feel. It's how most Americans feel. But here's what worries me. The people who should know something about AI, who should know the most about AI, they seem to be pretty perplexed by the whole thing too. By that I mean the people in the government who are responsible for regulating this new industry, or the people at investment banks who are giving out loans to AI startup companies. Or how about the people who are actually building AI, the agents, the bots who don't fully comprehend and understand what they themselves are creating. That's really what worries me. And I'm beginning to sense that a lot of them are struggling to get their own minds around what it is they've unleashed in the world. One sign of that is that anthropic, one of the biggest AI companies, you may have heard this recently brought in some theologians, some ethicists, some philosophers to help them get their brains around what are the ethical implications of what it is they're creating and how should they corral it. And it feels a little bit like I'm grateful they did that, but it feels a little bit like maybe it's a tell that they themselves don't even have a full understanding of what's going on. So it feels like we're kind of flying blind with this new power. And the rhetoric I'm hearing is that if this AI experiment is actually successful, it could fundamentally change human civilization and in the worst case scenario, may even end human civilization. But on the flip side, if this AI experiment we're all engaged in right now proves to be unsuccessful, well, then it could crash the entire global economy, because so much is riding on it, and lead us into a new global depression. Apart from AI's complexity, the other part of this whole thing that's been really difficult for me to get my head around is just all the different facets to this issue. First of all, there's the economic dimension. Most of us have heard stories about how AI is threatening to displace workers, the Pope released his first encyclical this year dealing with the dynamics of AI on workers and how it could replace people in the economy in a way that we haven't seen since the automation of labor intensive jobs. But now it's threatening jobs like mine. More creative jobs. What does that mean? There's a. A podcast I listen to with some regularity on economics and business and there's a lot of chatter going on right now about the potential for an AI bubble that could sink the entire economy. So there's the economic dimensions of this, then there's the political dimensions. There's growing political backlash to the building of AI data centers around the country. Ezra Klein just did a whole podcast episode about this at the New York Times. And there's the backlash to the big tech giants and the tech companies that have seemingly endless control over American politics and policies. So there's the political dimension to this story, there's the relational dimension to it. There's a fear that AI, as it becomes increasingly indistinguishable from actual human beings, may change the way we interact with each other, may change basic social dynamics, the development of social interaction for a whole new generation of young people. We're dealing with the fact that AI companies are quickly, and this is so creepy, moving into the area of pornography, where on the one hand you go, well, maybe it's a good thing because it could replace actual human victims of trafficking and exploitation with just AI generated non human stuff. On the flip side, AI generated content is so quick to proliferate that it could accelerate the malformation of a new generation of people with terribly graphic content that still dehumanizes people and commoditizes sex. That's not good. China is actually curtailing AI because they're worried that if it becomes too relational, it would have a negative impact on the birth rate in that country where they're already struggling with not having a growing population. There's the legal aspects of AI. Who's responsible under the law? When an AI agent advises somebody to take their own life or coaches someone how to take someone else's life, can our legal system deal with who's responsible when accidents are caused by autonomous vehicles or the harm caused by deep fake videos and just spirals on and on and on? The Supreme Court can't even keep up with what's going on with AI. And of course, there's the spiritual dimension to all of this. How does this technology provoke questions about what does it really mean to be human? How do Christian ethics relate to the use of these things. That's, again, what the Pope's encyclical was about. And let's be honest, at least within the evangelical church, there's not a great track record of being discerning about how to employ technology for good. The evangelical world seems so enamored with every new thing that comes down the pike that they just employ it. There's already talk about creating AI pastors, and that doesn't even get into the proliferation of plagiarizing sermons, which has been done forever but now has accelerated even more with AI. So the rabbit hole in AI just goes deeper and deeper and deeper. And I've wanted to talk about this issue on this show for a while, but it's been extremely difficult to find a good guest, a good conversation partner around this because it's changing so quickly, because there are so many facets and because there's so few people who can actually get their brains around this topic. But I'm really happy to report that that's changed. And my guest today is the best person I've encountered so far to talk about this. Megan Sullivan is a professor of philosophy at the University of Notre Dame, where she directs the university's Institute for Ethics and the Common Good. She's become one of the most prominent figures bridging Christian theology and AI ethics, and in September of her team hosted a summit about AI at Notre Dame, where they officially presented what's called Delta. It's a new framework for navigating AI in a responsible way. The Delta acronym stands for dignity, embodiment, love, transcendence and agency. Sullivan led the effort at the summit's gathering as the keynote speaker, and she led the effort for framing out this Delta architecture. Unlike economists and tech leaders who say that Christians shouldn't be too worried about ethics, but just want to optimize AI as a technology for automation and for economic expansion, Sullivan says that Christians, of all people, should be most concerned with human dignity. And she notes that a lot of the discussions right now about AI ethics tends to hover around what she calls the ethical floor, meaning basic principles of safety and fairness. And those things matter, but they're not sufficient given the power of this technology and its potential. So she says that we need to elevate the conversation to be about human flourishing and dignity. And as one of the leading voices on ethics in AI, she's been part of meetings at the Vatican. She was among those theologians, philosophers and ethicists that were brought in by Anthropic a couple months ago. Her team's leadership in this area is why last year, Notre Dame and her Institute of Ethics and Common good received a $50 million grant from the Lilly foundation specifically to further implement that Delta approach to AI. I am, for all those reasons, super grateful that she came on the show. But before we jump into the episode, I have to confess that when I first recorded this conversation with Megan Sullivan, I thought we were going to use it on the flagship Holy Post podcast show. And that's why the interview's a little bit shorter than my normal long, hour long or hour plus long conversations here on the Skypod. So I'm sorry the interview isn't longer, but here's the good news. I. I think this is just the first of many future conversations I'm going to be having with Megan Sullivan on this show and perhaps other Holy Post media outlet shows. Okay, here is my conversation with Megan Sullivan.
- [08:14]
B
She calls herself the Claude Mother, like the godmother of Claude.
- [08:33]
A
Megan Sullivan, welcome to the Holy Post.
- [08:36]
B
Oh my gosh, it's so great to be with you.
- [08:38]
A
I've followed your work, I've been intrigued by what you're doing, and I've wanted somebody on the show forever who can talk intelligently and experientially about this whole AI phenomenon. And you are top of the list, so I'm thrilled to have you. I want to begin with your visit to the Anthropic headquarters back in March. You were part of a group of philosophers, theologians, faith leaders. It got a lot of publicity that Anthropic was bringing folks like you in to help them think about the AI models that they're creating. And one of the things you wrote in one of your substack posts was that one of the worst mistakes that these LLMs, these large language models are making right now is giving these models personalities, these that mimic human personalities. Why is that so problematic? And what about your experience at Anthropic? Kind of made you realize this is maybe a bigger problem than people realize?
- [09:39]
B
Yeah. One kudos to Anthropic. I've been working on this issue pretty intensely for the last few years with a lot of my colleagues in the Christian world. And we have been to meetings and convenings where tech companies have sent representatives. But for me at least, this was the first time that a tech company had invited us to their headquarters, kind of in the sanctum sanctorum where Claude is being built and trained to really try to engage in this crucial conversation about what this technology is and what kind of world we're going to create with it. So I appreciated that that was a risk that Anthropic took, really trying to build that bridge out to the Christian community. And they've done other meetings like this over the last few months. And I think it really shows how important this kind of leadership is. When I it's funny, I was on NPR the week before I went on the Anthropic trip and I was being interviewed and I said super confidently, you know what we should do? We should just wipe the personalities out of these large language models. Just make them behave more like what we recognize as computer programs and not flatter us, not seem to have personality or emotion the way Claude or Gemini or Grok does. Grok has a terrible disposition. Shocking, shockingly, that a lot of these models also mimic their personalities of their founders. But I just said confidently, let's just remove that element from these models that we're now using in schools and workplaces. I said that so confidently, not really having a deep sense of how this technology is built and how these models are trained. When I was at Anthropic and got to spend some time with their extraordinary leadership, who are very generous with those of us that don't have PhDs in neuroscience or machine learning, they were like, let's think about this for a second. First, these models are built on language. They're built on just these vast troves of language based data that are all over the Internet. And you can't extract emotion and personality and character from language. Actually, it's the stuff that our language is made of. So one just conceptually the idea of having a large language model or like a language powered artificial intelligence that doesn't have some remnants of some kind of character or emotion, that's actually very hard. It's like saying you want water, but it's never wet. It's denying its kind of essence. But what Anthropic spend a lot of time talking with us about is this puzzle that they face. They can't get the kind of emotion and personality out of a language model, but they can direct the kind of characteristics that the model has. How it behaves, is it polite to you or is it mean to you? Does it push back against your ideas or is it sycophantic? Does it flatter you? Do we want our powerful super AI butlers that we're being promised from these tech companies to be funny and irreverent or to be like super servile? These are all questions that these tech companies are desperately trying to answer right now because the success of their product Depends on it having the types of characteristics that people will want to engage with. But also, this was very interesting about the conversations with the Anthropic leaders. There's a safety question built into this puzzle of what kind of personalities we train into. We train around these AI'. If your super powerful artificial agent is willing to obey your every command, like really willing to treat you as it's God and you command it to steal my bank data or teach you how to make a weapon, we might think, man, the AI should not accompany some of the worst aspects of you, the human user as well. If we make the AI really negative or really aggressive, kind of like a bully to its users, then we can imagine exacerbating some of the really terrible mental health trends that we've been seeing with LLMs. So there's this genuine puzzle that these tech companies are dealing with and it's a puzzle Christians are going to have a whole lot to say about. It's kind of the oldest question in the Christian tradition. It's a question God had in the garden of how do you have this agent that's capable of so much? How do you prepare it with the kinds of characteristics that's going to cause it to not just be powerful, but to be good? And that's why Anthropic, it's like a business proposition for them right now to talk to these philosophers and theologians because they're now asking the same question that drove Plato and Moses crazy.
- [14:58]
A
Okay, this opens up a whole bunch of questions for me. Part of it is related to the Pope's encyclical that relates to AI, and he mentions in there that there's a danger of having this kind of technology in the control of market forces which really drive towards efficiency, control, obviously profit. So when Anthropic or any other AI company brings you in as a consultant or these theologians, these ethicists, these philosophers, and they start talking about how do we create models that have personality but are virtuous, that will genuinely help their human operators but not enable evil? Those are really big questions. But at the end of the day, it has to be profitable. It's a little bit like saying, I want to build a casino and that casino needs to be ethical in how it makes money. But I don't want to be unprofitable. So, you know, I'm going to use all the tricks in my toolbox. Like there's going to be no windows, there's going to be no clocks, we're going to pump extra oxygen in the room. We're going to use all the different things. So when it comes to AI, isn't there a tension at some point between the ethics that you and your colleagues might be presenting to these companies and the need to be as profitable as possible? Recent reporting is showing that these AI companies have just enormous, enormous debt and a lot of pressure on them to turn things around and become more profitable. Is that when these ethical questions get thrown out the window?
- [16:41]
B
Of course, I don't know if they get thrown out the window, but the market pressures are real dynamic. Saw this Chris Ola, one of the co founders of Anthropic, he was at the Vatican when Pope Leo released that encyclical. And in his statement afterwards, he said, world, we in the industry are under such profound economic pressure to rush this to market and it's going to take other institutions, the church, the government, other kinds of forces that are capable of at least reining in and directing economic energies to weigh in now to try to make sure that this doesn't go off the rails. Because I think we can all agree that just trying to find a product that is going to, in the short term, make as much money as possible is not a recipe for building a product that's going to be healthier, sustainable for humanity. And we were lucky in the nuclear era in like the 1940s that the capacity to build atomic bombs never really, you know, other than maybe a couple like weapons, the military industrial complex, for most of us we were not able to apply economic pressures either way. So it's super interesting. I mean I'm a bit, I am definitely not an economic nihilist. I don't know where you stand on this, but I think the economic forces are powerful. But like any kind of powerful force, they're also unpredictable. And there's oftentimes powers that the mainstream media are not noticing. One thing I see with many of these large tech companies is they're trying to bring a product to market. Capitalism dictates that they will succeed or fail based on whether or not users adopt their product and pay for it, whether or not individuals buy subscriptions, whether companies pay for licenses. And those individuals and organizations are not, we're not mindless zombies. Like, we don't just be like, oh, Google has marketed, marketed this to me or Anthropic has marketed this to me. So I'm immediately, I must, must obey. You know, we have consciences, we have choices, we have choice. In the AI market right now, you know, there are seven big for big frontier models, at least in the United States that are jockeying for position, I think they are starting to cotton on to the fact that if you're going to build this technology that's going to be so ingratiated in people's lives, it better reflect or respond to their values, otherwise they're gonna reject it. And if they reject it, your company's toast.
- [19:25]
A
Yeah, okay, maybe. I don't know if I'm an economic nihilist, but we also know from human history that sometimes people have a proclivity towards things which are not for their benefit.
- [19:38]
B
What are you saying?
- [19:39]
A
I know it's hard to believe because
- [19:40]
B
I drink my fourth Coke Zero of the day.
- [19:43]
A
So for that reason, we collectively get together as families, as congregations, as church churches, as communities, and ultimately as whole societies to decide.
- [19:53]
B
You know what?
- [19:54]
A
For our own good, we need to regulate some of these things and put boundaries on them. So certain drugs can be fine at small doses, but at higher doses, you need a prescription, you need a doctor that's going to give that to you. Now, there's a legal market for those substances, but that's shunned upon, and we have legal structures that limit them. Similarly, I could imagine there being certain AIs that may have addictive qualities or really harmful qualities that people would nonetheless want if they could get access to it. So who is regulating that? Who is saying this is what the public should be allowed to engage? And here's what. The public should not be allowed to engage because it's too dangerous, either individually or communally. And are we seeing especially democracies function in a way that is putting regulation? Because it seems like what I'm hearing out of the current administration is essentially, go for it, AI companies. Do whatever you want. We're not going to stop you. I have a hard time believing the church is going to be able to do that.
- [20:56]
B
It's funny, because the main message we're hearing, at least in our country right now, is let her rip. Let's just see what happens when these companies just compete to the death. But occasionally you'll have these moments where even the federal government takes a pause in a deep breath. So when Anthropic had its controlled release of Methos a couple months ago, and unbeknownst to us, every single bank quietly had to update their software over the weekend because this AI found all kinds of exploitable bugs in just commercial banking software that we all thought was super safe until this one product hit. Then the administration was like, okay, whoa, whoa, whoa, whoa, whoa. Maybe we should take a pause. Maybe we should have some review before these things get released out into the wild. Because we realize a powerful, uncontrolled artificial intelligence could have the capacity to disrupt institutions and structures that we all just assume are safe and going to behave a certain way. So you do see these moments with the government of thinking like, oh, gosh, maybe we need a little bit more on what's going on, but not. But not. But. But not quite the will to. To do what it takes to regulate it. Obviously, the church, you know, as an entity, doesn't have the power to like, send AI executives to jail if they do something bad with their AI or to. To find them. We have a more, but we don't. Those who don't have any power. And again, the future of this technology depends in part on whether there's widespread adoption and how people adopt it. And a lot of folks right now are turning to their faith and turning to church communities to provide that kind of depth and insight and how to navigate a world that's changing really quickly. And church leaders, the kinds of. The places where we're formed, which I think are churches and schools. These places give folks some guidance and direction and grounding into what they should want. To your point, what's worth desiring and what's worth being careful about? And I think we might play a role, maybe not like soft regulators, but as formators, kind of preparing people to have the wisdom to make those decisions for themselves in the absence of the government making those decisions on our behalf.
- [23:46]
A
Okay, that raises an interesting topic that I'm glad we're going to have time to get into here. So I have a hypothesis, and it's not really mine, but it's an adaptation of someone else's. I read a book some years ago called how the west really Lost God, A New Theory of Secularization by a woman named Mary Eberstadt. And I'm sure there's a lot of scholars who've taken issue with her point of view, but the argument of her book was essentially secularization hypothesis has been as societies become more and more wealthy, people don't tend to be as religious. When they're less religious, they don't get married as frequently and they have fewer children. And that's kind of the sequence of things. And she argues that it's actually out of order that as societies become wealthier, people marry less and have fewer children. And that's what leads them to be less religious and gives all this evidence that it's when people end up being responsible for children, especially their own children, that it Raises all kinds of dilemmas for them, like, what do I really. How do I teach this new human being what's right and wrong? How do I answer their questions about death? How do I think about purpose and meaning? And all these bigger questions. And a lot of the. The stress that a community or a family or an individual feels around raising children is what leads them to religion and religious institutions for help. Because these are institutions that have been working for many cases, millennia on forming human beings in how to function well. So here's my hypothesis. In a weird way, is this whole AI moment a time when modern civilization is essentially having a baby, and we are now facing the dilemma of how do I raise this machine child virtuously? Which explains why Silicon Valley is calling in theologians, philosophers, and ethicists and asking these essentially religious questions. And could this whole dilemma of how do we form virtuous AI in a weird way lead us back to God as a civilization? The way people who suddenly are responsible for raising a child or maybe haven't thought about religious questions in the past, and now they are because they're putting their little one to bed and she's asking, hey, what happens when the dog dies?
- [26:13]
B
Interesting. So will this lead us back to, oh, my God, I have so much to say on this. So, all right, if I lose one of the points, you got to bring me back. But here's what I want to talk about in order, okay? Is making AI like raising a machine baby. Will this lead us back to God and just understanding why the rise in our economic productivity and capacity seem to correspond with our falling away from God. I want to start with maybe that third point, okay. Because I read this article in the newspaper yesterday that gave the New York Times, gave me such a headache. I think this woman is right. But what she was pointing out made me feel really sad. She said that she'd come from a first gen family, had gone to Harvard, had been just a real striver her entire life, had a poster of Sheryl Sandberg in her dorm room that said she's like, lean in, go after it, don't be afraid. And she was talking about how now she's having a midlife crisis and she realizes she doesn't really want to work, she just wants to be rich. And she's talking to her friends. She would just prefer to be rich and to actually not have to work. And she was talking about talking with other, like, young, younger folks, like emerging adults, and pointing out evidence that this is a trend. Like a lot of college students 10 years ago said. Like college students 40 years ago said the biggest thing they want in their life is a family. You know, 10, 15 years ago, the biggest thing they want is like an important job. A job that's going to enable them to be important in society. And now they're like, don't worry, this
- [27:59]
A
is not the end of the episode. There's actually plenty more. But to listen to the rest, you need to be a Holy Post plus subscriber. So head over to HolyPost.com SkyPod and sign up. For just $5 a month, not only will you get uninterrupted episodes of the Sky Pod, which means you'll never have to hear this dumb announcement again, but you'll also get access to everything else at Holy Post plus, including episodes of Getting Schooled by Kaitlyn Chess, bonus interviews, live streams, the Holy Post Book Club, exclusive merchandise, and a whole bunch more. And you'll get the warm fuzzy feeling of knowing that you're supporting our work of creating smart, pro neighbor Christian content. So head over to holypost.com skypod and subscribe today.