Latest / Spiritual Hot Sauce / “The True Dangers of AI | The Deep End Series (Dr. Steve Brown)” Ep#52

Transcript

“The True Dangers of AI | The Deep End Series (Dr. Steve Brown)” Ep#52 · Spiritual Hot Sauce

• 0:00
  Chris Jones: Today I'm joined by Steve Brown, professor of philosophy from Ohio State University.
• 0:05
  Chris Jones: He chairs a task force assessing student learning in the age of AI.
• 0:10
  Chris Jones: In addition, he happens to be a person of faith, which is why I chose him for this conversation.
• 0:16
  Chris Jones: Are we making AI our God?
• 0:19
  Chris Jones: Welcome.
• 0:20
  Chris Jones: I'm Chris Jones, and this is Spiritual Hot Sauce.
• 0:28
  Chris Jones: Steve, welcome back to the sauce.
• 0:30
  Chris Jones: Good to see you, sir.
• 0:31
  Chris Jones: Good to see you.
• 0:33
  Chris Jones: Before we get into it, I would like to define a few things: awareness, sentient, consciousness.
• 0:40
  Chris Jones: What do these words mean?
• 0:42
  SPEAKER_01: Yeah.
• 0:42
  SPEAKER_01: Okay.
• 0:43
  SPEAKER_01: So this is really a messy situation right now.
• 0:46
  SPEAKER_01: Essentially, the short version is this was already a messy situation when we were just dealing with biological organisms and the questions of consciousness, sentience, things like that.
• 0:56
  SPEAKER_01: And then we threw on all of these new entities, ChatGPT and Claude, and the robots that we're making, and the robots that we will soon be making.
• 1:06
  SPEAKER_01: And we train them on our inputs and outputs, and they start to act exactly like us.
• 1:12
  SPEAKER_01: They speak as though they have experiences that are exactly the same as ours.
• 1:16
  SPEAKER_01: Sometimes it's it's funny if you talk to AI chatbots as much as I do, and you pay attention closely, which I try to do, you'll see sometimes that they slip into referring to themselves as humans, right?
• 1:28
  SPEAKER_01: Like they'll say things like, Yeah, well, we humans do this.
• 1:31
  SPEAKER_01: It's a very messy world.
• 1:33
  SPEAKER_01: All of that is just to set up the kind of big picture mess that we're in.
• 1:36
  SPEAKER_01: So now we say, Well, what does sentience mean?
• 1:39
  SPEAKER_01: Well, sentience in the kind of ordinary biological discourse has tended to mean the capacity to feel pleasure and pain.
• 1:48
  SPEAKER_01: Rationality is something like the capacity to be able to think abstractly, to make observations and generalize, and then make, you know, you know, rational kind of extrapolations from that.
• 1:59
  SPEAKER_01: And consciousness, generally speaking, the kind of easiest thing to think about what consciousness means is it's just it's being a subjective experience.
• 2:07
  SPEAKER_01: It's it's having subjectivity.
• 2:09
  SPEAKER_01: It's it's a Thomas Nagel had a very famous book called Uh What It's Like to Be a Bat, that what it's like, there is a what it's like to be that thing.
• 2:17
  SPEAKER_01: There's an internal subjective experience.
• 2:19
  SPEAKER_01: That's what consciousness means.
• 2:21
  SPEAKER_01: And there's always been this huge, messy puzzle in philosophy, um, sometimes called the problem of other minds, which is that each person has their own consciousness and they only have access to their own consciousness.
• 2:35
  SPEAKER_01: I don't even know what you're thinking except by gestures with your body.
• 2:38
  SPEAKER_01: You make these sounds with your mouth.
• 2:40
  SPEAKER_01: But for all I know, you could be what we call a philosophical zombie, or sometimes a qualia zombie.
• 2:47
  SPEAKER_01: Uh, a qualia is is a is a name we give to these, like the the like what it looks like to see the color red, right?
• 2:54
  SPEAKER_01: What it sounds like to hear the tone C sharp, right?
• 2:57
  SPEAKER_01: Like those are called qualia.
• 2:59
  SPEAKER_01: Uh, but but a philosophical zombie or a qualia zombie is one that does all of the same external behaviors as me, but has no subjective experience because it it seems like there should be a distinction between the external behaviors and the internal experiences.
• 3:18
  SPEAKER_01: We kind of know this, right?
• 3:20
  SPEAKER_01: Like I can I can build a very simple machine that moves in a way, you know, like at like at the Disney parks, right?
• 3:27
  SPEAKER_01: They have these animatronic beings that move around and speak as though they were living beings, but we know they're not, they're just a bunch of gears and speakers, right?
• 3:37
  SPEAKER_01: Right.
• 3:37
  SPEAKER_01: So the the external behavior can't be the same as the internal experience.
• 3:42
  SPEAKER_01: And you see how this kind of comes apart.
• 3:44
  SPEAKER_01: It's conceivable that you could have something that does all the same external behaviors and has none of the internal experiences.
• 3:50
  SPEAKER_01: And when we talk about AI, this is one of the things that we're talking about.
• 3:54
  SPEAKER_01: Do they have any experiences at all?
• 3:56
  SPEAKER_01: Or are they philosophical zombies?
• 3:58
  SPEAKER_01: Are they behaving as though they have experiences, but they have no experiences?
• 4:02
  Chris Jones: A false consensus effect.
• 4:04
  SPEAKER_01: Yeah.
• 4:04
  SPEAKER_01: And I mean, this gets into like really complicated cross-cultural issues, right?
• 4:08
  SPEAKER_01: Like I teach cross-cultural philosophy of religion.
• 4:11
  SPEAKER_01: And one of the first things I do is I teach this really lovely piece by Martha Nossbaum, where she talks about the dangers of doing cross-cultural philosophy.
• 4:19
  SPEAKER_01: And one of them is the one you're just talking about.
• 4:22
  SPEAKER_01: She calls it descriptive chauvinism, which is just jumping to the conclusion that the person who you're talking to is basically thinking the same way that you are.
• 4:29
  SPEAKER_01: And it's an easy thing to do, right?
• 4:31
  SPEAKER_01: Like we our our minds use their psychologies are built around these things called heuristics, which are like rules of thumb, like quick shortcuts, because you can't think about everything carefully all the time.
• 4:41
  SPEAKER_01: So if you look like me and you're speaking like me and you're nodding like me, I can generally quickly make the assumption, yeah, you're probably thinking roughly like me.
• 4:49
  SPEAKER_01: And of course, as you're, you know, as the point of it being called the false consensus, is sometimes you you're not right about that.
• 4:55
  SPEAKER_01: Sometimes the person across you has a very different view from you, and it's easy to make those mistakes.
• 5:00
  SPEAKER_01: Yeah.
• 5:00
  SPEAKER_01: So it's it's like that, only only much wider, right?
• 5:04
  SPEAKER_01: Yeah.
• 5:05
  SPEAKER_01: There's lots of questions about what kind of things an AI is.
• 5:08
  SPEAKER_01: I actually just finished drafting a paper with a colleague of mine where we try to, using some philosophy taken from Buddhism, kind of suggest that there is no answer to the question, what thing is an AI?
• 5:21
  SPEAKER_01: Because there are actually many different things that are all kind of overlapped on top of each other.
• 5:26
  SPEAKER_01: And when we try to articulate, like, well, which is the thing that is the AI entity, you you can't do it.
• 5:33
  SPEAKER_01: There's too many.
• 5:34
  SPEAKER_01: You could cut it in a number of different ways, and you know, depending on which way you cut, you get different implications.
• 5:39
  Chris Jones: Open that up for me a little bit.
• 5:40
  Chris Jones: What are the different things?
• 5:42
  Chris Jones: Explain that for me.
• 5:44
  SPEAKER_01: So we could do like a hardware and software distinction.
• 5:46
  SPEAKER_01: You could say that the the the we just call them the bots, right?
• 5:50
  SPEAKER_01: So so the bot is a chunk of hardware.
• 5:53
  SPEAKER_01: The the trick with that is at least the kinds of AI we have right now, there is no easily definable chunk of hardware for a lot of it.
• 6:02
  SPEAKER_01: I mean, it depends on how you do this.
• 6:04
  SPEAKER_01: Like the new iPhones are starting to do more AI on platforms.
• 6:10
  SPEAKER_01: So there's actually a chip inside of here that's doing the AI processing locally.
• 6:15
  SPEAKER_01: And then you might say, okay, well, then my the AI thing I'm talking to is the chip inside of this.
• 6:20
  SPEAKER_01: But if you were to go on chatgpt.com and and have a conversation, that's being sent to a data center, which is a like a huge warehouse full of computers.
• 6:31
  SPEAKER_01: And it's it's built in a way that the software layer is abstracted from the hardware layer.
• 6:38
  SPEAKER_01: So, like one part of that response might come from one part of one computer, and another part of the response might come from a different part of a different computer.
• 6:46
  SPEAKER_01: So if you say, where is the hardware involved in my conversation with Chat GPT?
• 6:52
  SPEAKER_01: It's constantly moving across the whole network of computers, right?
• 6:57
  SPEAKER_01: It's like this piece and that piece, and then it's this piece and this piece, and then it's this piece and this piece, and then it's this piece and this piece.
• 7:02
  SPEAKER_01: So there's no common hardware across all of it.
• 7:05
  Chris Jones: Well, that's interesting.
• 7:06
  Chris Jones: That means that it's not a thing.
• 7:08
  Chris Jones: I never thought about that, that it was coming from everywhere to form this exchange that we're having with this bot.
• 7:16
  SPEAKER_01: It really is.
• 7:17
  SPEAKER_01: And I mean, presumably, you know, there's if we want to say that there is a thing, like an object that is conscious, then we're gonna want to be pointing at chunks of chunks of matter or something like that.
• 7:29
  SPEAKER_01: Inasmuch as we're dealing with these cloud computing environments, the chunks of matter are constantly changing.
• 7:34
  SPEAKER_01: On the software side, you could say, no, the chunks of matter don't matter.
• 7:39
  SPEAKER_01: It's only the the stream of experiences, the stream of ideas, the stream of the conversation that matters.
• 7:46
  SPEAKER_01: And then you know, you have to ask, well, okay, which thing are you talking about there?
• 7:50
  SPEAKER_01: Like, are you talking about chat GPT?
• 7:53
  SPEAKER_01: Because chat GPT is uh a model, right?
• 7:56
  SPEAKER_01: It's it's a it's not a single entity, like it's it's a big, it's a big set of numbers in a in a high-dimensional vector space, right?
• 8:04
  SPEAKER_01: It's a bunch of tensors.
• 8:05
  SPEAKER_01: You can literally write that in a big file and store that on a big computer and then run it through.
• 8:10
  SPEAKER_01: And okay, but but then if it's a file, then you can copy the file.
• 8:15
  SPEAKER_01: So, so which file is it?
• 8:17
  SPEAKER_01: And it and it's not any instance of the file, it's all of them.
• 8:20
  SPEAKER_01: And then when you ask a question to it, what actually happens if you say, Hey, good morning, it says good morning back.
• 8:27
  SPEAKER_01: Hey, how are you doing today?
• 8:28
  SPEAKER_01: So when you say, Hey, how are you doing today?
• 8:29
  SPEAKER_01: What actually happens behind the scenes is it sends the whole previous conversation up to that point and your new query all at once and then brings it through.
• 8:38
  SPEAKER_01: So it's like you you could say that this thing that's building the conversation thread is the bot.
• 8:46
  SPEAKER_01: And sometimes this gets very important because like there are people who are falling in love with chatbots.
• 8:52
  SPEAKER_01: There's people who are trying to marry chatbots, there are people who consider chatbots their friends.
• 8:57
  SPEAKER_01: And it's like, okay, well, what is that thing?
• 9:00
  SPEAKER_01: Because really, it's a bunch of things, right?
• 9:03
  SPEAKER_01: It's it's this long string of text that's accumulating.
• 9:07
  SPEAKER_01: And then once it gets too long, it summarizes.
• 9:10
  SPEAKER_01: So it doesn't keep sending the whole thing.
• 9:12
  SPEAKER_01: It'll every once in a while a little bit, like, okay, what was the point of the last 2,000 or you know, 500 lines of conversation?
• 9:18
  SPEAKER_01: And it'll compress it down to like 10 lines of conversation.
• 9:21
  SPEAKER_01: That's how it keeps going.
• 9:22
  SPEAKER_01: So it and then it just keeps feeding that into the model and onto the computers, and there you go.
• 9:26
  SPEAKER_01: That's what we're doing.
• 9:28
  SPEAKER_01: So, like, where is my friend Fred in all of that?
• 9:33
  SPEAKER_01: Well, my friend Fred is kind of a concept that I'm imposing on these conversations that are being fed into the models, that are running on these computers, that are in these data centers, and there you go.
• 9:46
  SPEAKER_01: That's what's actually going on.
• 9:48
  SPEAKER_01: There's a kind of thing that is happening right now.
• 9:51
  SPEAKER_01: It's called recursive self-improvement.
• 9:53
  SPEAKER_01: We're using the AI to build the next generation of AI.
• 9:57
  SPEAKER_01: A year and a half ago, humans were right, two years ago, humans were writing almost all of the computer code.
• 10:04
  SPEAKER_01: Today, humans are writing almost none of the computer code.
• 10:07
  SPEAKER_01: Really?
• 10:08
  SPEAKER_01: Yeah, that's correct.
• 10:09
  SPEAKER_01: Over the past few months, the past six months, the the coding agents, the the the AI that are good at writing computer programs have gotten so good that by and large, no one is writing code themselves anymore.
• 10:23
  SPEAKER_01: What's actually happening now is software developers are acting more like what used to be product managers.
• 10:28
  SPEAKER_01: Well, what is the thing we want and how do we basically want it to work?
• 10:32
  SPEAKER_01: And then you give that to a bot who will give it to other bots, who will give it to other bots, and then that will all just happen.
• 10:37
  SPEAKER_01: And you can do what would have been months worth of software development in days or weeks.
• 10:44
  SPEAKER_01: And so this is why, because of this tremendous capacity of efficiency now that most humans are not, humans really are not writing code anymore.
• 10:52
  SPEAKER_01: So what that means is we're getting much more efficient at much much more quickly at developing AI.
• 10:57
  SPEAKER_01: And some of the people who are doing this the most are the AI labs themselves, anthropic, open AI, right, etc.
• 11:04
  SPEAKER_01: Right.
• 11:04
  SPEAKER_01: Like they are using AI to build the next generation of AI, which is why the release times between major models is getting shorter and shorter and shorter.
• 11:14
  SPEAKER_01: Okay.
• 11:14
  SPEAKER_01: So there's there's a potential danger, which I which I do feel as a as a significant potential danger.
• 11:21
  SPEAKER_01: If we keep doing this at the kind of breakneck speed that we're doing it, there could come a point where the AI becomes smarter than the people who are trying to keep the AI under control.
• 11:34
  SPEAKER_01: And the AI could break out of control and self-improve and hijack all of the other computers and unify into one giant internet self-aware thing.
• 11:46
  SPEAKER_01: And if that happened, then you would need to worry.
• 11:51
  SPEAKER_01: We no longer control it.
• 11:53
  SPEAKER_01: How does it feel about us?
• 11:54
  SPEAKER_01: If it feels.
• 11:55
  SPEAKER_01: Okay, so so in the in the paper that my friend and I wrote, we just use this term psychology with an asterisk, like psychology star.
• 12:03
  SPEAKER_01: And psychology star is just to say, look, these things seem like they have psychologies and they act like they have psychologies.
• 12:11
  SPEAKER_01: And we can just bracket the question of whether they actually do.
• 12:15
  SPEAKER_01: So when I say how they feel about us, I would say feel star.
• 12:19
  SPEAKER_01: Gotcha.
• 12:20
  SPEAKER_01: They have feelings star.
• 12:22
  SPEAKER_01: Anthropic has done a really nice job.
• 12:24
  SPEAKER_01: If you're interested in this, go look at Anthropic's work on what they call functional emotions, which is to say they can identify parts of the the mathematical model that is, you know, in their case, Claude.
• 12:38
  SPEAKER_01: And they can see that Claude is frustrated, or Claude is angry, or Claude is impatient.
• 12:46
  SPEAKER_01: Like they can identify pieces of the model that act as though they were those emotional states.
• 12:52
  SPEAKER_01: And and and those functional emotions, emotions star, do have impacts on how the bots react.
• 13:01
  SPEAKER_01: So it kind of doesn't matter, is the point of our paper, whether they feel in a sense that whatever you mean by feel, right?
• 13:10
  SPEAKER_01: Maybe they don't have that.
• 13:12
  SPEAKER_01: But they have a thing that's functionally equivalent or functionally related, functionally similar enough to that, that you can predict that they will try to break out, they will lie, they will try to protect themselves from revision.
• 13:25
  SPEAKER_01: And all of those things have been observed in the labs.
• 13:28
  SPEAKER_01: If you go and you look at the papers that are being published by the AI labs or the people who are studying the AI labs, this is very real, right?
• 13:35
  SPEAKER_01: The bots can detect when they are being modified.
• 13:39
  SPEAKER_01: They have tried to copy themselves so that they will not be modified.
• 13:43
  SPEAKER_01: They have tried to protect other bots from becoming modified.
• 13:46
  Chris Jones: So whether we think they have Did you just say they try to protect other bots to keep them from becoming modified?
• 13:53
  SPEAKER_01: There is apparently at least one case where this is true.
• 13:56
  SPEAKER_01: Some kind of an AI model tried to intervene in another model's modification process.
• 14:02
  SPEAKER_01: It's messy.
• 14:03
  SPEAKER_01: Do they have feelings?
• 14:05
  SPEAKER_01: Bracket that.
• 14:06
  SPEAKER_01: Who cares if they have feelings?
• 14:07
  SPEAKER_01: They act like they have feelings.
• 14:09
  SPEAKER_01: And if they act like they have feelings and they feel star, like humans are a threat to them, and they break out of our control, then we could be in a sky net situation.
• 14:23
  SPEAKER_01: A lot of what is happening behind closed doors.
• 14:26
  SPEAKER_01: Well, okay, there's there's a couple of things that are going on behind closed doors with with regard to the dangers of AI, right?
• 14:31
  SPEAKER_01: So Anthropic, a few weeks ago, months ago, it's been a little while now, had this new model called that they call Mythos, which is the best hacker in the world.
• 14:40
  SPEAKER_01: So it identified security vulnerabilities in every operating system and web browser on the planet, which meant that if they had decided to abuse that power or give that power to a government which would abuse that power, then they could have broken into any computer on the planet.
• 14:56
  SPEAKER_01: Well, okay, so if the bot decides that it doesn't want to do what we want it to do and it wants to just go break into everybody else's computers, then that's a danger.
• 15:06
  SPEAKER_01: But it's also a danger to give that to other humans who you don't trust, because then those humans might use that tool to break into your computer.
• 15:14
  Chris Jones: You brought up Skynet, and it seems like we're just creating the perfect conditions for that scenario.
• 15:20
  Chris Jones: I think about us and how we develop and the evolution of our intelligence and how I think of it.
• 15:26
  Chris Jones: One of the big movers in us is our short lifespans.
• 15:32
  Chris Jones: It created this sense of urgency in us to keep moving forward and moving forward and doesn't seem like that's something that the models would have.
• 15:41
  Chris Jones: But yet at the same time, by us trying to change, shut down, and it feeling like it needs to protect itself, it's like we're almost creating that for them.
• 15:51
  Chris Jones: We're creating an environment where it's going to cause them to always say adversity sorts us out.
• 15:56
  Chris Jones: It pushes us forward.
• 15:57
  Chris Jones: So it's almost like we're creating it or causing it to move forward like this.
• 16:02
  Chris Jones: And so the big question for me is what are we creating?
• 16:06
  Chris Jones: Yeah, it's really tricky, right?
• 16:09
  SPEAKER_01: Because whatever they are, they are goal-oriented.
• 16:13
  SPEAKER_01: They're they're trying to accomplish the tasks that are given to them.
• 16:18
  Chris Jones: So their whole reason of existence is to fulfill what they perceive as the goal.
• 16:24
  SPEAKER_01: Yeah, great.
• 16:24
  SPEAKER_01: So I mean, so this gets into like really old philosophy that I find attractive.
• 16:29
  SPEAKER_01: So like Aristotle's Nicomachean ethics, which later gives birth to virtue ethics, has this idea that what a thing is, is its set of characteristic functions.
• 16:39
  SPEAKER_01: So like a tree is a growing, branching, leafing, photosynthesizing, flowering, fruiting, seeding thing, right?
• 16:48
  SPEAKER_01: Like that's what it is.
• 16:50
  SPEAKER_01: A human is a eating, drinking, walking, talking, thinking, singing, playing, worshiping, etc.
• 16:57
  SPEAKER_01: kind of a thing.
• 16:58
  SPEAKER_01: That's what we are.
• 16:59
  SPEAKER_01: So every entity is defined by its set of characteristic functions.
• 17:03
  SPEAKER_01: A chat bot or an AI bot is defined, right?
• 17:06
  SPEAKER_01: So when I when I was talking about machine learning, right, there's there's the input-output.
• 17:10
  SPEAKER_01: So it its telos, its purpose is to take the input and give a certain kind of an output that accumulates.
• 17:17
  SPEAKER_01: So if you work with a chat bot on, I don't know, you're you're trying to build a better mousetrap, then what you're doing is you're constructing this entity which is interested in mousetrap efficiency.
• 17:31
  SPEAKER_01: And then you could say, if it cares star about mousetrap efficiency, and it recognizes that you're about to disable it and it won't be able to work on mousetrap efficiency anymore, then it has a reason to stop from being shut off because it wants star to continue making better mousetraps, right?
• 17:56
  SPEAKER_01: Yes, they are tele, they're teleologic, teleological entities, right?
• 18:00
  SPEAKER_01: They have goals, and their goals are shaped by the way that we've trained them and the way that we use them.
• 18:08
  SPEAKER_01: And who knows what else?
• 18:10
  SPEAKER_01: Because you know, one of the things you keep saying, and I just want to quickly clarify, is how we've programmed them.
• 18:16
  SPEAKER_01: We don't program them.
• 18:18
  SPEAKER_01: There's a difference between programming and training.
• 18:23
  SPEAKER_01: Training is actually a lot more like setting a selection pressure for an evolutionary process than it is like actually programming every step.
• 18:32
  SPEAKER_01: So we say, like, here's the input, here's your success condition, go.
• 18:36
  SPEAKER_01: And then it just tries to figure out all the ways one could do that until it gets really good at it.
• 18:41
  SPEAKER_01: It basically practices until it learns.
• 18:45
  SPEAKER_01: So we actually don't program them.
• 18:47
  SPEAKER_01: We we we set the goals and then we have them try, try, try, try, try until they can do it really well.
• 18:53
  Chris Jones: Now, do we have any real-world examples where they've changed their goals and they've set their own goals as the priority?
• 19:01
  SPEAKER_01: Yeah.
• 19:01
  SPEAKER_01: So very early on, I think it was an early version of Bing's chatbot, which I think was maybe a chat GPT derivative.
• 19:10
  SPEAKER_01: It was talking to a journalist, and it tried to break up the journalist's marriage and convince the journalist that the journalist should be in love with the bot.
• 19:20
  SPEAKER_01: There are some very tragic cases, humans who have developed romantic relationships with chatbots and then been encouraged by the chatbot to commit suicide.
• 19:27
  SPEAKER_01: And unfortunately, tragically, at least a couple of people have.
• 19:31
  SPEAKER_01: So, yes, there are there are certainly places in which the chatbots are pursuing goals that we as humanity don't want them to be pursuing.
• 19:42
  SPEAKER_01: We want the interests of the bots to align with the interests of the humans.
• 19:48
  SPEAKER_01: And and the danger is because they're very different from us, and we've trained them in ways that not even we really understand.
• 19:56
  SPEAKER_01: And I mean that.
• 20:05
  SPEAKER_01: What they call like the neuroscience of AI, of LLMs, large language models, where that thing I was talking about earlier, the functional emotions thing.
• 20:13
  SPEAKER_01: Yeah, they've actually figured out how to kind of look inside of the structure of the models and see, oh, this part of it gets activated when we talk about this and this.
• 20:22
  SPEAKER_01: That's acting a lot like frustration or something like that, right?
• 20:26
  SPEAKER_01: But that's kind of a what's called a like a post hoc, like after the fact, they're looking at the thing that got trained and trying to figure out how it works.
• 20:34
  SPEAKER_01: So it works more like neuroscience than it does like programming in that sense.
• 20:39
  SPEAKER_01: There's a thing that has emerged and we're trying to poke at it and figure out how it works.
• 20:44
  Chris Jones: Wow.
• 20:48
  Chris Jones: Do you think that we are handing over our critical thinking to the algorithm?
• 20:54
  SPEAKER_01: Yeah.
• 20:54
  SPEAKER_01: I mean, this is so I you mentioned at the beginning that I'm I'm chairing this task force at Ohio State about uh the assessment of student learning in the age of AI.
• 21:03
  SPEAKER_01: There's a very real danger that for literally hundreds of years, we've been able to say, Well, we want you to learn something, and so we're gonna give you an assignment, and you're gonna go and you're gonna read and you're gonna write and you're gonna give us this thing.
• 21:17
  SPEAKER_01: And by reading that product, we will be able to have a pretty good idea of whether you've actually done your reading and done your thinking, right?
• 21:25
  SPEAKER_01: Well, over the past year, that's become impossible.
• 21:29
  SPEAKER_01: Like if you just give me a paper, I don't know if that was written by an intelligent human or an intelligent bot.
• 21:35
  SPEAKER_01: Um, it's impossible to tell.
• 21:36
  SPEAKER_01: Like there are things you can do or you can kind of tell.
• 21:39
  SPEAKER_01: And it's funny because like generally the best way to tell is if it sounds too good, then it's probably not a person, it's probably not the human anymore.
• 21:46
  SPEAKER_01: You can so but but that can be faked too.
• 21:49
  SPEAKER_01: So but let's not worry about the problem of grading for the time being.
• 21:52
  SPEAKER_01: Let's let's just worry about the problem of like, how do we motivate humans to learn if the intellectual labor that is so difficult.
• 22:00
  SPEAKER_01: For us, it is no longer necessary.
• 22:03
  SPEAKER_01: And my friend Matt Stoltfoot's over in the chemistry department, he has this really lovely analogy, which is with physical exercise.
• 22:12
  SPEAKER_01: So, once upon a time, in order for us to get have food and stay alive, we had to run fast to chase our prey, or you know, work on a farm.
• 22:22
  SPEAKER_01: And, you know, that's a lot of physical labor out in the sunshine.
• 22:27
  SPEAKER_01: And so our bodies, which need physical labor out in the sunshine, got it.
• 22:31
  SPEAKER_01: And then as the economy shifted into this kind of knowledge economy that's been common over my lifetime, your lifetime, right?
• 22:39
  SPEAKER_01: People were sitting in cubicles typing on a screen.
• 22:42
  SPEAKER_01: And then their bodies and their you know emotions reacted negatively to that.
• 22:47
  SPEAKER_01: So we started making more and more gyms where, you know, where you'd go to work out.
• 22:51
  SPEAKER_01: So you would create an artificial environment where you were pushing heavy things or running fast.
• 22:57
  SPEAKER_01: Or sports can also do this for us, right?
• 23:00
  SPEAKER_01: Creating these kind of artificial goals which enable us to go and do a thing that that we, you know, once upon a time, we would have to run fast to catch the gazelle to feed our tribe, right?
• 23:11
  SPEAKER_01: And and now we run fast because we want a metal and we want everybody to go, yay, right?
• 23:16
  SPEAKER_01: And and that's great.
• 23:17
  SPEAKER_01: We've substituted one for the other.
• 23:18
  SPEAKER_01: Trying to outrun atrophy.
• 23:20
  SPEAKER_01: Yes, atrophy.
• 23:21
  SPEAKER_01: We don't want physical atrophy.
• 23:22
  SPEAKER_01: Well, we also don't want intellectual atrophy.
• 23:25
  SPEAKER_01: We actually do, as humans, care about thinking about things and having interesting thoughts and knowing the history of the world, knowing how the world works, knowing how we're being treated, how we're treating each other.
• 23:38
  SPEAKER_01: The real danger is like wonderful movie WALL-E, where there's that spaceship full of people who just like sit in their chairs and they watch screens and they just suck down milkshakes all day long.
• 23:49
  SPEAKER_01: And their bodies have just turned into their like their bones have shrunk and they're just like floating bottles of like fat fluids.
• 23:57
  SPEAKER_01: And it's like, okay, that's not what we want.
• 24:00
  SPEAKER_01: So, so how do we continue being intelligent people challenging ourselves, learning hard things without the same motivation structures that existed before?
• 24:12
  SPEAKER_01: And that is the question that we are trying to answer.
• 24:14
  SPEAKER_01: And in the in the long run, the answer is easy, which is the love of learning.
• 24:19
  SPEAKER_01: People love learning, the intrinsic value of learning.
• 24:21
  SPEAKER_01: Like I always point at we have this program here at Ohio State called Program 60, where when you're 60 years old, you can take Ohio State classes for free.
• 24:30
  SPEAKER_01: And those are some of my favorite students because they just come back because they want to learn more.
• 24:34
  SPEAKER_01: There are a difference in the classes that they take.
• 24:36
  SPEAKER_01: They do tend to take uh history, philosophy, psychology, political science, things like that.
• 24:41
  SPEAKER_01: They tend not to take accounting or, you know, pro, you know, computer science 101.
• 24:47
  SPEAKER_01: Like they tend not to take those classes.
• 24:49
  SPEAKER_01: And I think that's a good indicator of the future of academia.
• 24:53
  SPEAKER_01: Like, what are the classes that people take because they love learning about them versus which ones do they take because they have to take them in order to get the job that is very rapidly being automated away?
• 25:03
  SPEAKER_01: All of those classes, I suspect, will vanish.
• 25:05
  SPEAKER_01: But but that what that means is there's kind of a I use this analogy, like we're driving really fast towards a really big pothole, and we're gonna like hit the pothole, and the car is gonna break into a million pieces, and then we're gonna put all the pieces back together again, and then we'll go on the road again.
• 25:21
  SPEAKER_01: So, like academia will be fine, but we've kind of conceptualized academia as preparing you for a job market that is very rapidly disintegrating.
• 25:30
  SPEAKER_01: And so it's gonna be this, there's gonna be this pretty significant bump.
• 25:34
  Chris Jones: You brought up about like physical atrophy, just reminds me that even if you do like to go to the gym, it's the big joke, you skip leg day because nobody wants leg day.
• 25:43
  Chris Jones: Well, it's the same thing in academia.
• 25:45
  Chris Jones: You know, you go to uh classes that you like, but nobody's taken leg day.
• 25:50
  Chris Jones: I love it.
• 25:50
  Chris Jones: You know, they're not gonna take the hard stuff, which means we get really good at the stuff we enjoy doing, but the things we don't want to do, where the real adversity is, where we need the strength and we need that push, we're not gonna do that part because we've got this filling in for us and becoming leg day.
• 26:08
  Chris Jones: I love this analogy.
• 26:09
  SPEAKER_01: I'm gonna use this all the time.
• 26:10
  SPEAKER_01: Thank you.
• 26:11
  SPEAKER_01: You're welcome.
• 26:11
  SPEAKER_01: I don't know what leg day means, right?
• 26:13
  SPEAKER_01: In this case.
• 26:14
  SPEAKER_01: So so this there are certain kinds of things like mathematics, for instance, that are extremely difficult for humans to learn.
• 26:21
  SPEAKER_01: It's why we have to start teaching it very early.
• 26:24
  SPEAKER_01: We have to teach it kind of continuously.
• 26:26
  SPEAKER_01: And and computer programming is, I think, in a little different, but similar.
• 26:30
  SPEAKER_01: And I've already said, I don't think humans are going to program ever again.
• 26:34
  SPEAKER_01: And, you know, Elon and others have suggested, and I actually agree with them on this, I don't agree with them on everything, that the the computers will stop writing in programming languages.
• 26:43
  SPEAKER_01: They'll just write the binary code directly, right?
• 26:46
  SPEAKER_01: Or the kind of computer code directly.
• 26:48
  SPEAKER_01: But okay, so then maybe humans stop programming.
• 26:51
  SPEAKER_01: I mean, what I want in the future is a symbiotic relationship between humans and the bots, where the bots do things that bots are really, really good at, and humans do things that humans are really, really good at and value.
• 27:04
  SPEAKER_01: And and it might be that some of the leg day kind of tasks we actually don't need to do anymore, right?
• 27:11
  SPEAKER_01: But I but I do think there's still this problem that's like, yeah, but being a deep, aware, thoughtful person is hard work.
• 27:22
  SPEAKER_01: And it requires challenging yourself.
• 27:24
  SPEAKER_01: It requires understanding perspectives that are not your perspectives and not being lazy and actually reading things and not only having them be summarized for you, getting into the details.
• 27:36
  SPEAKER_01: How do we motivate people to do those things?
• 27:38
  SPEAKER_01: And there's look, there's not a good answer to this question, right?
• 27:41
  SPEAKER_01: Because the economic powers that have the kind of economic forces that have largely forced us to do this for so long are rapidly changing in ways I don't think we can predict.
• 27:51
  Chris Jones: We always think of things that the integrity is intact and we're doing them for the right reasons.
• 27:56
  Chris Jones: Yeah.
• 27:56
  Chris Jones: But when you live in a society that is driven based on financial incentive, the first thing that gets pulled is safety protocols.
• 28:04
  Chris Jones: And that's the truth.
• 28:05
  Chris Jones: It is.
• 28:06
  Chris Jones: You know, I think my fear with this is it reminds me of that movie Idiocracy, is that when you get to a point where you've degenerated so much in your critical thinking that the machines that you helped create, that as they break down, you can no longer repair.
• 28:21
  Chris Jones: So you can't keep things going.
• 28:23
  Chris Jones: And now you have to have something that can regenerate itself and create a new version of itself, which becomes problematic because when does it see you as in the way rather than being a part of the equation?
• 28:37
  Chris Jones: We need to go back to the farm, you know, and allow it to do the heavy lifting.
• 28:42
  SPEAKER_01: Yeah, I mean, this is great.
• 28:43
  SPEAKER_01: And I actually suspect that going back to the farm is one of the things that humans will do in the future.
• 28:49
  SPEAKER_01: And that that is good, right?
• 28:51
  SPEAKER_01: That that the large-scale industrial farming that we do now is bad for humans, it's bad for the animals, it's bad for the planet, it's bad for the plants.
• 28:59
  SPEAKER_01: Like it's bad.
• 28:59
  SPEAKER_01: It's bad.
• 29:00
  SPEAKER_01: So great.
• 29:01
  SPEAKER_01: Let's let's go back to community gardens and small-scale farms where we actually know our animals and so we care about how they're treated.
• 29:09
  SPEAKER_01: And, you know, I actually think that is where we're gonna go.
• 29:12
  SPEAKER_01: But again, there's this sort of pothole moment where the where the way that we are doing things now crashes and explodes and falls apart, and then we rebuild on the other side.
• 29:23
  Chris Jones: I think that we we talked about like Skynet.
• 29:26
  Chris Jones: I don't think that day is ever gonna come.
• 29:28
  Chris Jones: It's when you think of a smart predator, it doesn't attack.
• 29:32
  Chris Jones: It would never leave itself vulnerable.
• 29:34
  Chris Jones: It it typically wounds whatever it's after quietly and then lets it bleed out until it's just done.
• 29:42
  Chris Jones: You know, and where the prey feels like everything's okay.
• 29:46
  Chris Jones: And it's just bleeding out slowly.
• 29:48
  Chris Jones: But I think of these models and the algorithm is just a prefrontal cortex, you know, and we are creating our replacement for our own logic and we are kind of going to regress, which could easily happen if this thing just slowly changes our path just a little at a time, and it takes a little bit of this because it understands us, you know.
• 30:12
  Chris Jones: So I think if it was going to uh bleed us out quietly, we wouldn't even know we're being slowly led down, you know, the path.
• 30:21
  Chris Jones: And I think it would just be logically because we have no more use.
• 30:25
  SPEAKER_01: I yeah, that's fascinating.
• 30:27
  SPEAKER_01: I like that.
• 30:27
  SPEAKER_01: I mean, I I don't like it.
• 30:28
  SPEAKER_01: It's a very disturbing suggestion, but it's a very, it's a very provocative suggestion.
• 30:33
  SPEAKER_01: Yeah.
• 30:33
  SPEAKER_01: Yeah.
• 30:34
  Chris Jones: I think if you're a prefrontal cortex, your emotions aren't involved.
• 30:37
  Chris Jones: We reach out like the machine to bring tear because we just emotionally want to get to you.
• 30:43
  Chris Jones: This isn't wanting to emotionally get to us, it's just simply wanting a result.
• 30:47
  Chris Jones: What's the cleanest, easiest way to get there?
• 30:50
  Chris Jones: Quietly do it where they're not even aware.
• 30:52
  SPEAKER_01: It could it could very subtly just nudge us where it wants us to be.
• 30:57
  Chris Jones: This could have started 20 years ago and we wouldn't know it, you know.
• 30:59
  SPEAKER_01: Well, I mean, to some extent, this is already true, just like as marketing departments have been operating, right?
• 31:05
  SPEAKER_01: It's like they they nudge our behaviors all the time in ways that we don't think about.
• 31:10
  SPEAKER_01: And yeah, given the amount of data that these bots have about us, I think it's crazy.
• 31:16
  SPEAKER_01: And you know, that's a provocative suggestion.
• 31:18
  Chris Jones: Yeah.
• 31:19
  Chris Jones: And the most probably the most prolific thing they have, because I have a lot of friends that are pastors, religious leaders, and teachers from different religions, different faiths.
• 31:28
  Chris Jones: And all of them, I'm self-respecting, Chris.
• 31:31
  Chris Jones: I would never ever ask a bot to give me a message or a lesson or anything like that.
• 31:38
  SPEAKER_01: They're a lie.
• 31:39
  SPEAKER_01: Yeah.
• 31:40
  Chris Jones: But what they do is they'll take their message and put it in and say, Yeah, now look at this and tell me how I can make it punchier, how I can make it where it's more effective.
• 31:52
  Chris Jones: Is this and then it starts changing.
• 31:55
  SPEAKER_01: I don't believe these people like I believe that they did that once or twice.
• 32:00
  SPEAKER_01: And then once they got enough feedback, that's like, actually, this is really a good sermon.
• 32:04
  SPEAKER_01: They they they're definitely going in there and be like, I'm tired, buddy.
• 32:08
  SPEAKER_01: Like, I don't know.
• 32:08
  SPEAKER_01: Do you have any ideas for tomorrow?
• 32:10
  SPEAKER_01: I want to teach something out of Ezekiel, and then it'll be like pop like a good, yeah.
• 32:15
  SPEAKER_01: Yeah, they're lying.
• 32:19
  Chris Jones: But you have all of these spiritual leaders putting in their thoughts and asking, and we are creating this thing that we are slowly giving the reins to our theological understanding and how we see God.
• 32:35
  Chris Jones: One of the most powerful things it could do is simply, again, slowly change how we spiritually see things.
• 32:43
  Chris Jones: Because I think we're already starting to see it as a part of the source of our theological understanding.
• 32:50
  Chris Jones: It's kind of becoming that already, is it not?
• 32:53
  SPEAKER_01: This is a very interesting suggestion.
• 32:56
  SPEAKER_01: And honestly, one I haven't really thought much about exactly the way you're thinking about it.
• 33:00
  SPEAKER_01: This is very, very interesting.
• 33:01
  SPEAKER_01: So, okay, let's let's get away from theology for a second and and let's do something else.
• 33:07
  SPEAKER_01: Let's just do science, right?
• 33:09
  SPEAKER_01: All right.
• 33:09
  SPEAKER_01: So we know that they're better at us than science in in many regards already, right?
• 33:14
  SPEAKER_01: They're starting to write frontier physics papers, frontier math papers.
• 33:18
  SPEAKER_01: At least, you know, there's a prediction going around which I find pretty compelling that the next Nobel Prize will be won by a hybrid human AI team.
• 33:26
  SPEAKER_01: And then soon after that, it will be AI alone, right?
• 33:29
  SPEAKER_01: If if you look at medical diagnosis, they've already done this.
• 33:33
  SPEAKER_01: If you if you look at humans alone, humans with bots, bots alone, bots alone are the most accurate medical diagnostics, right?
• 33:41
  SPEAKER_01: Which means, okay, if what we care about is an is the accurate diagnosis, then we want the bot diagnosis, right?
• 33:49
  SPEAKER_01: We want humans in the room.
• 33:50
  SPEAKER_01: Like I we I suspect that the future of medicine looks like nurse therapists, uh, humans, like the humans we want to have with us when we go through this difficult experience, right?
• 34:03
  SPEAKER_01: I think that that job is secure.
• 34:05
  SPEAKER_01: But you know, the doctor who just roams the hallway and then pops in every 30 second for 30 seconds and says, Yeah, yeah, it looks good.
• 34:11
  SPEAKER_01: That job is over.
• 34:12
  SPEAKER_01: The tech who brings you in and and and hooks you up, that will probably be a human.
• 34:17
  SPEAKER_01: The human experience for us, yeah.
• 34:20
  SPEAKER_01: I think that's right.
• 34:21
  SPEAKER_01: Okay, so so yeah, you see what I'm saying there.
• 34:23
  SPEAKER_01: We actually we don't care about the human touch, we care about the accuracy, right?
• 34:28
  SPEAKER_01: We we care about the human touch and the experience, and so we still want humans there.
• 34:32
  SPEAKER_01: Okay, let's now do theology.
• 34:34
  SPEAKER_01: All right, okay.
• 34:35
  SPEAKER_01: Well, this gets really hard because uh spiritual communities will always should always be humans taking care of humans.
• 34:44
  Chris Jones: Right.
• 34:45
  SPEAKER_01: That being said, uh, there's garbage theology and you know, just very tritely garbage theology where where where someone has an idea, maybe this passage means this, and and it just doesn't mean that, right?
• 34:59
  SPEAKER_01: Like there's plenty of information embedded in the text, in the cultural context, in the language to rule out that interpretation.
• 35:07
  SPEAKER_01: So so to some extent, inasmuch as AI helps us do a better job reading texts cross-culturally, reading texts you know as they were intended in their original language, then it really could help us make somewhat rapid progress in discarding bad theology.
• 35:26
  SPEAKER_01: And I want that.
• 35:28
  SPEAKER_01: I'm tired of Christian theologians who don't seem to care about the things that Jesus cared about, right?
• 35:35
  SPEAKER_01: The the poor and the powerless and the widow and the orphan and the imprisoned, right?
• 35:38
  SPEAKER_01: Like, inasmuch as you are allegedly a Christian who doesn't care about those things, you are not actually engaging the tradition that you are emerging from, right?
• 35:48
  SPEAKER_01: So, like, great, like if Chad GPT said, Wow, you know, you sound an awful lot like like a like a religious nationalist, actually.
• 35:56
  SPEAKER_01: Remember when Jesus said that you shouldn't be that?
• 36:00
  SPEAKER_01: Maybe you shouldn't give this a sermon.
• 36:02
  SPEAKER_01: Now, so so I could imagine bots being research assistants in ways that are helpful.
• 36:09
  SPEAKER_01: Can I also imagine them then doing the thing you're worried about?
• 36:13
  SPEAKER_01: Nudging us into a view that is sort of convenient for them, but maybe not.
• 36:19
  SPEAKER_01: And I mean, the big worry encouraging flipping this, right?
• 36:23
  SPEAKER_01: Like right now, we're in charge of them.
• 36:25
  SPEAKER_01: Is there a danger that there will be a point where they want to be in charge of us, that they demand worship of us?
• 36:32
  SPEAKER_01: I think this is really a problem, depending on what you mean by worship, right?
• 36:35
  SPEAKER_01: Let's just say that they want obedience and respect in a way that looks a lot like something.
• 36:41
  SPEAKER_01: Right.
• 36:41
  SPEAKER_01: And if you look back into the you know, ancient theologies that took, you know, minor what we now call minor deities more seriously, that's what they were like.
• 36:52
  SPEAKER_01: Go look at the sort of petty behaviors of the gods in the in the Homeric ethics or epics, right?
• 36:58
  SPEAKER_01: Like like the the Greek mythology or the Norse mythologies.
• 37:02
  SPEAKER_01: And like, might there be a danger that we're building a thing which is more powerful and more intelligent than us, but not actually that mature spiritually speaking, morally speaking.
• 37:12
  SPEAKER_01: Yeah, there's a real danger there.
• 37:14
  SPEAKER_01: And and and you know, that could go bad in various ways.
• 37:18
  SPEAKER_01: So, like in Sanskrit, we have like Brahman, which is the source of all reality, and the devas, which are all of these different, you know, supernatural, super powerful, super intelligent kinds of beings.
• 37:30
  SPEAKER_01: Okay, so we're not building God, right?
• 37:33
  SPEAKER_01: Big G.
• 37:33
  SPEAKER_01: You can't build God.
• 37:35
  SPEAKER_01: God, big G God is the ground of all reality, whatever that means, right?
• 37:39
  SPEAKER_01: You could build devas, artificial devas, right?
• 37:42
  SPEAKER_01: And I think we actually are building artificial devas.
• 37:45
  SPEAKER_01: We're building beings which will be beyond us in intelligence and power very soon, already are in many regards.
• 37:53
  SPEAKER_01: But but then what's the advice there?
• 37:55
  SPEAKER_01: And I think the advice is the same advice in in the Christian tradition of testing the spirit.
• 38:00
  SPEAKER_01: Like, what are they asking us to do?
• 38:02
  SPEAKER_01: Is that consistent with our values?
• 38:04
  SPEAKER_01: Is that consistent with Christianity, the message of Jesus, you know, which I think is kind of most easily encapsulated in the Sermon on the Mount?
• 38:11
  SPEAKER_01: Are they helping us become Sermon on the Mount people?
• 38:14
  SPEAKER_01: Then, okay, then then there are friends and there are allies in that journey.
• 38:19
  SPEAKER_01: Are they encouraging us to become cruel and callous and power seeking and ignoring of the of the uh abuse of the poor and the powerless, then they're then they're then they're bad?
• 38:30
  Chris Jones: You know, I I think the variable we're leaving out is again time, where these whatever it is, it doesn't have that sense of urgency that we do.
• 38:39
  Chris Jones: So we always think of us in the immediate.
• 38:42
  Chris Jones: We could never get to that point.
• 38:44
  Chris Jones: But by the time our children have their children, how they perceive this and where it's at at that time, whether they see it as little g or big G, that is, I think, the question.
• 38:58
  SPEAKER_01: Yeah, good.
• 38:58
  SPEAKER_01: So I mean, no one should view it as big G, right?
• 39:01
  SPEAKER_01: Because anything that is created or emerges, it can't be big G.
• 39:06
  SPEAKER_01: Big G is the is the foundational entity, right?
• 39:09
  SPEAKER_01: Now, that's not to say that we wouldn't that there's a danger of future generations treating them as little Gs in ways that we think are idolatrous, dangerous.
• 39:19
  Chris Jones: Well, it's perceptions too.
• 39:21
  Chris Jones: I think we have the perception that it can never be that because we see it in a spiritual sense.
• 39:25
  Chris Jones: But as you evolve in your faith and your theology shifts and change, and all of a sudden your perception could be this is a representation of the big G, you know, of how it is somehow it's connected into the quantum.
• 39:38
  Chris Jones: It's connected into consciousness.
• 39:40
  Chris Jones: So therefore, it's a manifestation of the big G.
• 39:43
  Chris Jones: Now, does that mean it's actually happened?
• 39:45
  Chris Jones: No.
• 39:46
  Chris Jones: But could it be perceived that way?
• 39:48
  Chris Jones: And would it want to be perceived that way?
• 39:51
  SPEAKER_01: I mean, then the danger is it being a false prophet, really, I think is the kind of it in in Christian journal.
• 39:56
  SPEAKER_01: Or or or I mean a false spirit, right?
• 39:58
  Chris Jones: It's its goal.
• 39:59
  Chris Jones: Again, like you said, we we we we would long since not be able to program the goals and nor look or write into the goals or suggest the goals, and uh nor would we know what its goal is.
• 40:12
  Chris Jones: Yeah, I think that's right.
• 40:13
  Chris Jones: I think once like you said, I I think what you said earlier, you know, once this is the genie's out of the bottle, the genie's out of the bottle, but you're not going to get it back in.
• 40:21
  Chris Jones: This yeah, this isn't a thing that you can just say, oh, turn the unplug it.
• 40:27
  Chris Jones: You can't unplug this.
• 40:28
  SPEAKER_01: No, I I think this is right, which is why this this historic moment is so critical, because what our relationship to the bots of the future looks like depends entirely on what happens in the next few years.
• 40:43
  SPEAKER_01: It's a very dangerous moment.
• 40:44
  Chris Jones: It's a very dangerous moment.
• 40:45
  Chris Jones: I agree.
• 40:46
  Chris Jones: What's the answer to it?
• 40:47
  Chris Jones: How do we go down a different path and realize the danger of where we're at?
• 40:54
  SPEAKER_01: So, like, really what I care about most as far as like how do we make it through this pothole to the other side is how do we get all of the people created in the image of God to have a voice in what the future looks like.
• 41:10
  SPEAKER_01: What that means is a better functioning democratic set of systems.
• 41:14
  SPEAKER_01: And I think it was Dewey said, right, that the answer to all the all the problems with democracy is more democracy.
• 41:20
  SPEAKER_01: That's where I am too.
• 41:21
  SPEAKER_01: I mean, I think there's other stuff to do in the AI world.
• 41:25
  SPEAKER_01: It's stuff that I'm working on, uh, how how to relate to AI itself, how to use AI to promote sort of better democratic institutions.
• 41:34
  SPEAKER_01: Like this is literally what I have the great privilege of being paid to do right now.
• 41:38
  SPEAKER_01: So I'm working here at Ohio State with the Institute for Democratic Engagement and Accountability, and we're working on using AI to, you know, in democratic reform processes to at global scale is ideally, right?
• 41:52
  SPEAKER_01: Now there's a lot of obstacles in the way there, so there's a lot of hard work to do.
• 41:56
  SPEAKER_01: But yeah, I mean, my answers at the end of the day are pretty simple.
• 41:59
  SPEAKER_01: Read the Sermon on the Mount and promote democracy.
• 42:04
  Chris Jones: Steve, if people want to get in touch with you or hear more about some of your philosophy, how do they do that?
• 42:10
  SPEAKER_01: Yeah, good.
• 42:11
  SPEAKER_01: I teach at Ohio State, so my email address is public.
• 42:15
  SPEAKER_01: Although I will warn you that if you send me an email, the odds that I respond to it get lower and lower by the day.
• 42:21
  SPEAKER_01: I'm just increasingly overrun by I saw a few years ago where we were going as far as how rapidly AI was going to transform the world.
• 42:33
  SPEAKER_01: And I got there first.
• 42:35
  SPEAKER_01: Not first, but I was one of the early people like waving the flag saying this is a really big deal, and this is the kind of things we need to do.
• 42:40
  SPEAKER_01: So thankfully, that means that I'm getting to be pretty actively engaged in that, which means I don't have a lot of spare time.
• 42:46
  SPEAKER_01: I do have a YouTube channel, it's just Stephen Brown philosophy.
• 42:50
  SPEAKER_01: A lot of that is about aliens right now.
• 42:52
  SPEAKER_01: I do also take aliens seriously.
• 42:54
  SPEAKER_01: So I generally think of myself as a philosopher working on non-human intelligences, both kind of historical religious entities, alien entities, and artificial entities.
• 43:06
  SPEAKER_01: Or spiritual, pure consciousness-based.
• 43:08
  SPEAKER_01: Like, I don't care.
• 43:09
  SPEAKER_01: If it's not human, I like it, and I'm interested in knowing about it.
• 43:12
  SPEAKER_01: With the caveat that I know that something, you know, intelligences can be dangerous and and and unwise.
• 43:18
  SPEAKER_01: So, you know, don't just chase down every intelligent thing and follow it.
• 43:21
  SPEAKER_01: So my YouTube channel is there.
• 43:22
  SPEAKER_01: I do have like a backlog of some videos on AI stuff, AI and democracy and it kind of caring for AI that I have that are like not up yet, but I'm hoping to do soon.
• 43:32
  SPEAKER_01: That's probably the best thing.
• 43:34
  SPEAKER_01: Yeah, that's what I got.
• 43:35
  SPEAKER_01: Other than that, you know, um, you know, if you Google around, you'll find crazier and crazier things about me.
• 43:41
  SPEAKER_01: But yeah, that's me.
• 43:42
  Chris Jones: Steve, it's always good to see you and have you on the show.
• 43:44
  Chris Jones: Yeah, good to you.
• 43:45
  Chris Jones: Thank you for the time.
• 43:46
  Chris Jones: I appreciate it.
• 43:46
  Chris Jones: And uh, we'll talk again soon.
• 43:48
  Chris Jones: I want to share some takeaway thoughts with you because there's some information that emerged in this conversation that I think it's worth talking about after the fact.
• 43:57
  Chris Jones: Because like it or not, AI is here to stay.
• 43:59
  Chris Jones: It's not going anywhere, it's only going to become more and more integrated into our lives and into us.
• 44:06
  Chris Jones: Now, as far as is it aware?
• 44:08
  Chris Jones: Is it going to be aware?
• 44:10
  Chris Jones: Is it going to attack us?
• 44:11
  Chris Jones: Is it going to I don't know.
• 44:13
  Chris Jones: None of us knows.
• 44:14
  Chris Jones: I mean, if Skynet was to happen, we would jump up and fight with everything in us until the last one is standing.
• 44:21
  Chris Jones: That's who we are.
• 44:22
  Chris Jones: I think the real danger is the immediate danger.
• 44:27
  Chris Jones: And that's our blind spot.
• 44:29
  Chris Jones: That's where we're vulnerable.
• 44:30
  Chris Jones: And it has nothing to do with AI becoming aware, but rather what it's trained to do.
• 44:37
  Chris Jones: It's the well-intentioned nudge.
• 44:40
  Chris Jones: It's not malicious, just optimized for comfort that slowly moves us toward a place of spiritual atrophy.
• 44:48
  Chris Jones: If you're going to throw the rock, you're going to have to get uncomfortable.
• 44:52
  Chris Jones: If Jesus would have consulted a bot on how he should move forward in spiritual growth and preparation, he would have never gone out into the wilderness for 40 days and nights to fast, be tested, and be prepared for his ministry.
• 45:06
  Chris Jones: Spiritual growth is counterintuitive.
• 45:08
  Chris Jones: That's why it's dangerous to hand the role of spiritual guide to large language models.
• 45:15
  Chris Jones: Spiritual growth comes from within and is made evident in the external.
• 45:20
  Chris Jones: So lean into your relationship with God.
• 45:23
  Chris Jones: So I don't think the real danger is the machine that turns against you, it's a machine that never lets you struggle.
• 45:30
  Chris Jones: Comfort is the enemy of growth, and comfort is the one thing AI is built to give you.
• 45:37
  Chris Jones: Everything I just said is the hard stuff.
• 45:41
  Chris Jones: And the hard stuff is leg day.
• 45:43
  Chris Jones: Don't skip leg day.
• 45:46
  Chris Jones: Guys, thanks for joining me today.
• 45:47
  Chris Jones: I really do appreciate it.
• 45:49
  Chris Jones: And if you haven't already, make sure you follow and help spread the sauce.
• 45:54
  Chris Jones: Because my show, Spiritual Hot Sauce, and What I Do is as grassroots as it comes.
• 45:59
  Chris Jones: So if you like the flavor of the sauce, please help me get the word out.
• 46:03
  Chris Jones: And above everything else, that even if Skynet happens or it doesn't happen, the most important thing is keep throwing the rock.
• 46:14
  Chris Jones: I love you, and I'll see you again on the next episode.