Latest / A New Angle / Part two: Crissie McMullan & Jason Wiener are AI Realists

Transcript

Part two: Crissie McMullan & Jason Wiener are AI Realists · A New Angle

0:00
[music]
0:07
SPEAKER_02: This is A New Angle, conversations about how we're doing and where we're going.
0:13
SPEAKER_02: I'm your host, Justin Angle.
0:15
SPEAKER_02: This show is supported by the Headwaters Foundation,
0:18
SPEAKER_02: First Security Bank, and the University of Montana College of Business.
0:22
[music]
0:30
SPEAKER_02: Hey, folks, welcome back, and thanks for tuning in.
0:32
SPEAKER_02: Today, we bring you part two of my conversation with Chrissy McMullen and Jason Weiner.
0:37
SPEAKER_02: Together,
0:38
SPEAKER_02: these two created AI Realists,
0:40
SPEAKER_02: a nonprofit organization aimed at helping communities shape AI policies.
0:45
SPEAKER_00: One of the things that I am actually enjoying the most about working on AI issues is
0:50
SPEAKER_00: that it really makes you go deep about everything.
0:56
SPEAKER_02: Let's pick it up with a discussion about their policy ideas.
1:00
SPEAKER_02: Chrissy, another thing you're working on is this idea of citizens' assemblies.
1:03
SPEAKER_02: What are those and how are you making them come to life?
1:06
SPEAKER_00: So citizens assemblies are a new-ish idea that have been gaining traction in places
1:14
SPEAKER_00: as far reaching as Paris to Kentucky.
1:18
SPEAKER_00: And what they are is a different way for us to have citizens involved in legislation.
1:26
SPEAKER_00: So in theory, our state legislature is a citizen legislature.
1:29
SPEAKER_00: We elect them and they are regular people who live among us and then they go
1:34
SPEAKER_00: to Helena Every other year and make decisions on our behalf.
1:38
SPEAKER_00: Another way to also have a decision or,
1:43
SPEAKER_00: you know,
1:43
SPEAKER_00: a recommending body that thinks about how we want to govern AI,
1:49
SPEAKER_00: for example, would be citizens assemblies that are drawn by lot.
1:54
SPEAKER_00: Okay.
2:09
SPEAKER_00: Think specifically and deeply about that issue.
2:11
SPEAKER_00: They bring in different experts.
2:13
SPEAKER_00: They read different materials.
2:14
SPEAKER_00: They can request different types of information and evidence.
2:18
SPEAKER_00: They have legal expertise on hand to help them think about what would actually be
2:23
SPEAKER_00: a legal solution that would be viable or not.
2:27
SPEAKER_00: And what's really exciting about the ones that have been tried out so far is
2:33
SPEAKER_00: that they often come up with really nuanced policy solutions.
2:37
SPEAKER_00: Okay.
2:38
SPEAKER_00: Because they don't have anything else.
2:40
SPEAKER_00: They're not trying to win any other kind of game.
2:42
SPEAKER_02: Yeah.
2:42
SPEAKER_02: They're not trying to win elections either.
2:44
SPEAKER_00: They're not worried about what... Yeah, exactly.
2:47
SPEAKER_00: They're not worried about the election that's coming up or they're not worried about
2:51
SPEAKER_00: the lobbyists who's got their ear.
2:53
SPEAKER_00: They're not worried about...
2:55
SPEAKER_00: They're just trying to do the best they can with the information they have and
2:58
SPEAKER_00: to support their families and communities.
3:01
SPEAKER_00: So it can also have this other indirect outcome of depolarizing issues
3:10
SPEAKER_00: because people aren't there to represent a party.
3:12
SPEAKER_00: They don't have party politics.
3:14
SPEAKER_00: They are just there to, again, come up with the best solutions they can for a thorny issue.
3:20
SPEAKER_02: In the places you've seen this done, how is it then interfacing with the policymaking process?
3:28
SPEAKER_00: Yeah, so I just read a book called Politics Without Politicians
3:31
SPEAKER_00: that really gives two examples from France.
3:35
SPEAKER_00: One of those was around climate change issues, and the other was around end-of-life issues.
3:42
SPEAKER_00: And they were both called into being by the president of France,
3:49
SPEAKER_00: who in the climate change one,
3:51
SPEAKER_00: he initially said, Okay, well, whatever you guys come up with, we will enact.
3:57
SPEAKER_00: He dialed that back a little bit over time, as politicians tend to do.
4:01
SPEAKER_00: But they did still have a direct,
4:03
SPEAKER_00: you know,
4:03
SPEAKER_00: basically,
4:04
SPEAKER_00: they came up with,
4:04
SPEAKER_00: I think,
4:05
SPEAKER_00: 140 policies that they gave to the French Parliament,
4:08
SPEAKER_00: and the French Parliament adopted a version of Almost all of them.
4:12
SPEAKER_00: Wow.
4:14
SPEAKER_00: And then in the end of life issues when they did something kind of similar,
4:18
SPEAKER_00: but they started off with a little less carte blanche.
4:22
SPEAKER_00: A few more constraints.
4:24
SPEAKER_00: Yeah, they had learned from that.
4:25
SPEAKER_00: And I think that's an important part of the Citizens Assembly idea is
4:30
SPEAKER_00: that it will be an experiment.
4:32
SPEAKER_00: We will have to...
4:33
SPEAKER_00: Figure things out as we go,
4:35
SPEAKER_00: but there is a really strong precedent that we can get stronger policies
4:40
SPEAKER_00: that are better reflective of what the people actually want.
4:44
SPEAKER_02: Speaking of what the people want, Jason,
4:46
SPEAKER_02: on your site,
4:47
SPEAKER_02: you have this polling exercise where people are expressing opinions,
4:52
SPEAKER_02: people are voting on those opinions and expressing new opinions.
4:55
SPEAKER_02: Talk about that and what are you learning?
4:56
SPEAKER_01: Yeah, this is an instance of a platform called Polis.
4:59
SPEAKER_01: Been around for over a decade,
5:01
SPEAKER_01: and it uses not the latest generation of AI,
5:05
SPEAKER_01: but some algorithms in order to identify clusters of opinion.
5:11
SPEAKER_01: And the way it works is you give people a series of statements across the range
5:16
SPEAKER_01: of opinions on an issue,
5:18
SPEAKER_01: and they say, agree, disagree, unsure, pass, kind of a thing.
5:22
SPEAKER_01: And then you can,
5:23
SPEAKER_01: over time,
5:24
SPEAKER_01: begin to see where people clump together, What people tend to agree with each other.
5:30
SPEAKER_01: What statements tend to identify agreement from people who otherwise disagree with each other.
5:36
SPEAKER_01: And you can sort of like map the public opinion landscape in
5:40
SPEAKER_01: a way that's like relatively easy to comprehend.
5:43
SPEAKER_01: People can contribute their own opinions for others to vote on
5:46
SPEAKER_01: if they don't feel like they're being well served.
5:49
SPEAKER_01: It's been used,
5:50
SPEAKER_01: Taiwan used it to good effect in deciding how to handle Uh,
5:55
SPEAKER_01: rideshare and regulate that industry alongside taxis and things like that.
6:00
SPEAKER_02: Anything in the data that you can tell a story about yet?
6:03
SPEAKER_01: I think what we're seeing so far is like a, a large degree of agreement around, uh,
6:11
SPEAKER_01: A concern about where this is headed.
6:14
SPEAKER_01: The most universally supported statements tend
6:17
SPEAKER_01: to be things around people should be compensated for the way
6:22
SPEAKER_01: that their work is training these models.
6:27
SPEAKER_01: And that kids should be protected from manipulation by chatbots and things like this.
6:33
SPEAKER_01: I would say it's a self-selecting group that's going to the website that's filling these out.
6:39
SPEAKER_01: I'd love to drive traffic from across the ideological spectrum, such as there is one on this.
6:47
SPEAKER_01: Because I think there's really no one who wants us to go extinct because of this.
6:52
SPEAKER_01: So this is just sort of like Version 0.1 of engaging with people.
6:58
SPEAKER_01: But just, again,
6:59
SPEAKER_01: trying to solicit more than,
7:00
SPEAKER_01: like, I identify with this group or identify with that group.
7:04
SPEAKER_01: But, you know, this is what I believe.
7:06
SPEAKER_01: And where can I find commonality around that?
7:09
SPEAKER_01: You know,
7:09
SPEAKER_01: our hope is to take some of these issues that really fall out with
7:13
SPEAKER_01: the greatest support and then put that back to the people who have engaged with us
7:16
SPEAKER_01: and the poll and say, you know, like, form an affinity group.
7:20
SPEAKER_01: Start to articulate this, like, You know, what do you want and who are you going to ask for it?
7:25
SPEAKER_01: Christy,
7:26
SPEAKER_01: are you learning anything
7:26
SPEAKER_02: about Montana's views on some of the infrastructure,
7:31
SPEAKER_02: like data centers and other things that sort of building up of the sector of
7:37
SPEAKER_02: the economy will have big effects on?
7:40
SPEAKER_00: Well, data centers are a hot topic right now, of course, across the state.
7:46
SPEAKER_00: Jason and I keep laughing because it's such a hot topic and there are
7:50
SPEAKER_00: so many really strong organizations,
7:52
SPEAKER_00: especially a lot of the environmental organizations,
7:54
SPEAKER_00: public interest groups,
7:55
SPEAKER_00: who are taking a lead on figuring out You know,
7:59
SPEAKER_00: what we need to do around data centers that Jason and I have kind of been like,
8:02
SPEAKER_00: okay, I think actually this is an area we don't need to step in.
8:05
SPEAKER_00: Okay.
8:05
SPEAKER_00: Other people are working on that.
8:07
SPEAKER_00: Yes.
8:07
SPEAKER_00: And it is interesting because even when we,
8:09
SPEAKER_00: you know,
8:09
SPEAKER_00: brought this up at the town hall and kind of explained our position on it,
8:13
SPEAKER_00: it was still one of the most popular, you know, topics.
8:17
SPEAKER_00: People really want to talk about data centers.
8:18
SPEAKER_00: Sure.
8:19
SPEAKER_02: The use of AI has a much larger resource footprint than the use of a Google search,
8:24
SPEAKER_02: for example.
8:24
SPEAKER_02: I don't know if that's still necessarily true, but for a time, it was quite a big difference.
8:30
SPEAKER_02: Yet when you're just typing a question into ChatGPT,
8:34
SPEAKER_02: the salience of that consumption effect isn't so high.
8:40
SPEAKER_02: Or if you see a data center,
8:42
SPEAKER_02: that is a pretty visceral representation of the resources
8:46
SPEAKER_02: that these systems are demanding.
8:49
SPEAKER_02: So I think it is an easy thing for people to hold onto,
8:52
SPEAKER_02: but I think there's some merit in that as well.
8:55
SPEAKER_01: Yeah.
8:55
SPEAKER_01: I mean, I think we should be smarter about how we generate electricity and
9:00
SPEAKER_01: then would be less concerned about its consumption.
9:02
SPEAKER_02: Yeah.
9:02
SPEAKER_02: And that is one argument that the AI proponents trot out there.
9:06
SPEAKER_02: It's like,
9:06
SPEAKER_02: this will force us,
9:07
SPEAKER_02: the energy demands are so great that this will force us
9:09
SPEAKER_02: to innovate our energy systems.
9:13
SPEAKER_02: It's a hard thing to review.
9:13
SPEAKER_02: Is that how it's worked out so far?
9:15
SPEAKER_02: Exactly.
9:15
SPEAKER_01: I mean, ask the people of Memphis who have like one gas turbine after another going in next
9:20
SPEAKER_01: to them.
9:20
SPEAKER_00: Yeah.
9:21
SPEAKER_00: I mean, and I think another issue around data centers that is
9:25
SPEAKER_00: a super incredibly important point is just the lack of transparency.
9:30
SPEAKER_00: Just the way that the corporations are going about citing these,
9:35
SPEAKER_00: placing these, trying to move these permits forward without any input into...
9:40
SPEAKER_00: How it's happening is it is kind of a great metaphor for the for AI in general.
9:45
SPEAKER_00: Right.
9:45
SPEAKER_00: It's just hard to it's impenetrable, as we talked about earlier.
9:50
SPEAKER_00: And so understandably, people are mad.
9:53
SPEAKER_00: They're like, I should know what's coming in my backyard.
9:57
SPEAKER_00: Especially if it's going to be emitting noise 24-7,
10:00
SPEAKER_00: if it's going to be taking my water,
10:02
SPEAKER_00: if it's going to be,
10:04
SPEAKER_00: you know,
10:05
SPEAKER_00: sending,
10:05
SPEAKER_00: you know,
10:06
SPEAKER_00: increasing the temperature of the groundwater that I rely on,
10:09
SPEAKER_00: you know, there's lots of reasons to be mad about it.
10:12
SPEAKER_00: So we're not saying that people shouldn't be mad at it.
10:14
SPEAKER_00: Basically, we're just like, this issue seems to be covered by a lot of really
10:17
SPEAKER_02: smart people.
10:19
SPEAKER_02: In our remaining time,
10:20
SPEAKER_02: I'd love to just sort of pull the lens out a little bit and just talk
10:24
SPEAKER_02: about some broader issues that are implicated in this conversation.
10:29
SPEAKER_02: Things like trust in a society, things like learning and education, social skills.
10:37
SPEAKER_02: Some of this might inspire your work with AI Realists,
10:40
SPEAKER_02: but how are you thinking about You mentioned your daughter at the head
10:44
SPEAKER_02: of this conversation, Chrissy.
10:46
SPEAKER_02: I think about this with my kids.
10:48
SPEAKER_02: My daughter expressed a ton of fear about AI getting in the way
10:54
SPEAKER_02: of her learning what she needed to learn.
10:55
SPEAKER_02: And this was in the context of a math class.
10:58
SPEAKER_02: And then she sort of figured out a way to get AI to coach her through a series
11:03
SPEAKER_02: of steps with some algebra problems, et cetera.
11:07
SPEAKER_02: And it seemed to be an effective use of the technology to foster learning.
11:12
SPEAKER_02: Yet I see it with my students and they're choosing a wide variety of ways to bypass thinking.
11:19
SPEAKER_02: How are you thinking about this with your own daughter?
11:21
SPEAKER_00: One of the things that I am actually enjoying the most about working on AI issues is
11:27
SPEAKER_00: that It really makes you go deep about everything.
11:33
SPEAKER_00: What is education?
11:35
SPEAKER_00: How do we learn?
11:36
SPEAKER_00: Why do we learn?
11:37
SPEAKER_00: What is important to learn?
11:41
SPEAKER_00: What is the kind of ultimate outcome you're trying to go for?
11:44
SPEAKER_00: And then it kind of keeps being this thread because you're like,
11:47
SPEAKER_00: well, if you're learning to get a job, what is the job?
11:52
SPEAKER_00: What jobs are going to be available?
11:53
SPEAKER_00: And so if you're not learning to get a job, what are you learning for?
11:57
SPEAKER_00: And it goes really deep.
11:58
SPEAKER_00: And, you know,
11:59
SPEAKER_00: in some ways,
11:59
SPEAKER_00: I think it's really interesting because it's like,
12:03
SPEAKER_00: oh, what is kind of the value of life itself, you know, and it's scary to realize.
12:12
SPEAKER_00: Revisit what's the point of life and what's the meaning of life for sometimes.
12:17
SPEAKER_00: But at the same time, it can be really exciting and interesting.
12:20
SPEAKER_00: And I think it's for me,
12:22
SPEAKER_00: it's been really rewarding because we as a society are starting
12:26
SPEAKER_00: to have these conversations.
12:27
SPEAKER_00: Yeah.
12:28
SPEAKER_02: Jason, I certainly don't know many folks that have sort of such a deep training
12:33
SPEAKER_02: in philosophy but also are so fluent in the technology itself.
12:39
SPEAKER_02: As you've kind of been a student of both and a practitioner of both,
12:42
SPEAKER_02: how are you thinking about the future of thinking and writing and our ability to ask
12:48
SPEAKER_02: and answer the questions that Chrissy was talking about a moment ago?
12:51
SPEAKER_01: There's something about interacting with a thing that is so fluent in language
12:56
SPEAKER_01: that makes you wonder about subjectivity,
12:59
SPEAKER_01: like the internal life that we all have,
13:02
SPEAKER_01: what it is to be, you know, a sentient thing in the world.
13:05
SPEAKER_01: Does this thing have an inner life?
13:08
SPEAKER_01: You know,
13:08
SPEAKER_01: there's like widely ranging opinions about that among people who,
13:12
SPEAKER_01: you know,
13:12
SPEAKER_01: my opinion should know better about like what the state of
13:14
SPEAKER_01: the current technology is.
13:16
SPEAKER_01: And so.
13:19
SPEAKER_01: Given this, it is an opportunity for us to reflect on what we want our own inner life to be.
13:28
SPEAKER_02: I think about human interaction and social skills,
13:32
SPEAKER_02: and we have so many off-ramps to having face-to-face conversations.
13:36
SPEAKER_02: Yet I juxtapose that against the massive rise in the popularity of podcasts.
13:43
SPEAKER_02: Somehow people love listening and or watching people talk.
13:48
SPEAKER_02: And the only way I can think of explaining that is it's gotta be some facsimile
13:52
SPEAKER_02: of social interaction that people are hungry for.
13:55
SPEAKER_02: If AI podcasts were gonna replace us all,
14:00
SPEAKER_02: you'd probably already see that happening,
14:03
SPEAKER_02: but people aren't tuning in to AI podcasts.
14:06
SPEAKER_02: And the ones that I've seen attempted,
14:08
SPEAKER_02: I think they have like a party trick attraction to them,
14:12
SPEAKER_02: but they're not,
14:14
SPEAKER_02: maybe the technology gets there,
14:15
SPEAKER_02: but people don't seem that interested
14:17
SPEAKER_01: Models are not good at generalizing out
14:19
SPEAKER_01: of distribution.
14:20
SPEAKER_01: You train them on a corpus of text and what they will do is more or less pick out
14:25
SPEAKER_01: the most probable bits of that text in order to regurgitate.
14:30
SPEAKER_01: A podcast should confront you with something new,
14:34
SPEAKER_01: interesting, something that you haven't thought of before.
14:38
SPEAKER_01: That's not what these models are great at.
14:42
SPEAKER_01: They're good at a lot of things.
14:44
SPEAKER_01: They're good at summarizing what's already out there.
14:47
SPEAKER_01: They're good at it.
14:48
SPEAKER_01: Crawling a huge corpus of data in order to distill some insight or maybe a plan
14:54
SPEAKER_01: of action,
14:55
SPEAKER_01: but they don't have the rich inner life that we have by almost any standard.
15:01
SPEAKER_01: And so we should seek to preserve our rich inner life in the face of this technology
15:07
SPEAKER_01: that has the potential to sort of make us more objects in the world,
15:11
SPEAKER_01: like things with descriptors that can be acted upon by forces,
15:14
SPEAKER_01: rather than cultivating the rich inner life that is Mm-hmm.
15:35
SPEAKER_02: I could probably create some sort of a bot in Claude to go to some fake gym
15:44
SPEAKER_02: and work out and then log the workouts so that they appear on Strava and
15:49
SPEAKER_02: that all my friends can see how I did whatever, but I didn't really do it.
15:54
SPEAKER_02: And nobody ever really wants to do that.
15:56
SPEAKER_02: I mean, people fake stuff for social presentation,
15:59
SPEAKER_02: but I'm never going to fake that I did a workout.
16:03
SPEAKER_02: Because it's more efficient to do that.
16:05
SPEAKER_02: Even if I could take a pill to simulate the effects,
16:08
SPEAKER_02: still there's something in the struggle of using my body.
16:12
SPEAKER_02: And whether you're using your body,
16:14
SPEAKER_02: your brain,
16:14
SPEAKER_02: or whatever faculty you have,
16:16
SPEAKER_02: I think there is something in that that is distinctly human that I can't imagine
16:20
SPEAKER_02: an AI taking that away from me.
16:22
SPEAKER_01: Albert Borgman,
16:23
SPEAKER_01: professor at the University of Montana,
16:25
SPEAKER_01: my teacher, he called this goods internal to a practice.
16:29
SPEAKER_01: And this is engagement.
16:31
SPEAKER_01: And AI can inform goods internal to a practice.
16:34
SPEAKER_01: I took a wordworking class.
16:36
SPEAKER_01: The project that I came up with And so these can definitely become scaffoldings
16:39
SPEAKER_01: to improve our lives and help us achieve our ends.
16:42
SPEAKER_01: But that is something that will require intention on our part as users.
16:59
SPEAKER_01: The engagement of our social support system and our political economy to make sure
17:04
SPEAKER_01: that what we're doing is cultivating humanity and not just sort of farming people.
17:10
SPEAKER_02: We'll be back to my conversation with Chrissy McMullen and Jason Weiner after this short break.
17:18
SPEAKER_02: A New Angle would not be possible without the generous support of
17:21
SPEAKER_02: the Headwaters Foundation,
17:23
SPEAKER_02: working side by side with Western Montanans to improve the health
17:26
SPEAKER_02: of our communities.
17:28
SPEAKER_02: And you can't do that sort of work without access to capital and education.
17:33
SPEAKER_02: Two critical catalysts offered by First Security Bank and the University
17:36
SPEAKER_02: of Montana College of Business.
17:39
SPEAKER_02: Hey, this is Mark Moss from Tell us Something, and you're listening to A New Angle.
17:46
SPEAKER_02: Welcome back to A New Angle.
17:48
SPEAKER_02: I'm speaking with Chrissy McMullin and Jason Weiner about AI realists.
17:53
SPEAKER_00: I definitely have wondered in the past,
17:54
SPEAKER_00: like, oh, are we going to have to go to mental gyms in the future?
17:57
SPEAKER_00: Are we going to have to, like,
17:58
SPEAKER_00: create an intentional, like,
18:00
SPEAKER_00: I'm going to think for at least an hour every day because,
18:03
SPEAKER_00: you know, we don't automatically have to think very much anymore.
18:07
SPEAKER_00: And even in white collar jobs,
18:08
SPEAKER_00: if our jobs are most,
18:10
SPEAKER_00: if our thinking is outsourced,
18:12
SPEAKER_00: we may still at some point realize, like, oh, wait, my brain is not healthy anymore.
18:16
SPEAKER_00: I need to go to the gym and it will feel better and I'll be able to think better
18:20
SPEAKER_00: and there'll be benefits.
18:22
SPEAKER_00: And we, you can kind of see an analogy to that in the world of physical fitness
18:27
SPEAKER_00: that there was,
18:29
SPEAKER_00: many of us used to have to work with our bodies and you wouldn't want to go to a gym
18:32
SPEAKER_00: because you already worked with your body for, you know, all day.
18:35
SPEAKER_00: Yeah, the idea of going to a gym is sort of absurd in and of itself.
18:38
SPEAKER_00: Yeah, you're right.
18:38
SPEAKER_00: Right, exactly.
18:39
SPEAKER_00: And then we had an obesity epidemic.
18:42
SPEAKER_00: And, you know,
18:42
SPEAKER_00: because it was no longer the default for people to work during the day,
18:47
SPEAKER_00: kids weren't, you know, playing outside during the day anymore.
18:50
SPEAKER_00: We weren't by default eating a more healthy diet.
18:56
SPEAKER_00: We were eating by default food that was not making us well.
19:00
SPEAKER_00: So I think it's a both and situation.
19:02
SPEAKER_00: And when you look at what are the defaults?
19:04
SPEAKER_00: So if the defaults with AI become that we outsource our intelligence,
19:10
SPEAKER_00: if we outsource thinking,
19:11
SPEAKER_00: if we outsource any little source of friction,
19:14
SPEAKER_00: we will probably see some kind of, at least statistically speaking, bad results.
19:20
SPEAKER_00: And some people will more quickly than others,
19:23
SPEAKER_00: because of advantages they have or education they already have,
19:27
SPEAKER_00: they'll start to realize, oh, wait, I have to keep engaging in the real world.
19:30
SPEAKER_00: I have to choose to have real relationships, even if they're harder.
19:33
SPEAKER_00: I have to choose to think.
19:36
SPEAKER_00: I have to read books, even if it's harder and less fun than just chatting with my chatbot.
19:41
SPEAKER_00: But I think the default may not be that way.
19:43
SPEAKER_00: And I think that that can bring down our society.
19:47
SPEAKER_00: Like at one point I was working on childhood obesity issues and like 30%
19:51
SPEAKER_00: of children were predicted to have type 2 diabetes by
19:54
SPEAKER_00: the time they graduated high school.
19:57
SPEAKER_00: So if you're like, oh,
19:58
SPEAKER_00: if 30% of our population is part of this default and experiencing the negative sides
20:04
SPEAKER_00: of a default of outsourcing our thinking, we're going to need to change that.
20:09
SPEAKER_02: I think that surfaces the,
20:11
SPEAKER_02: at least from my vantage point as a professor here in front of students,
20:15
SPEAKER_02: trying to prepare them for the world that they'll enter,
20:19
SPEAKER_02: hopefully as some form of a worker, maybe, or citizen, let's put it that way.
20:26
SPEAKER_02: It's this concept of discernment.
20:29
SPEAKER_02: So these AIs can be tremendously powerful to help them solve problems,
20:34
SPEAKER_02: to get work done, to be more efficient, to be more productive, whatever it is.
20:40
SPEAKER_02: Yet they can also bypass thinking,
20:42
SPEAKER_02: bypass socialization,
20:43
SPEAKER_02: bypass writing,
20:46
SPEAKER_02: all kinds of, there's off ramps to effort available wherever you look.
20:52
SPEAKER_02: And some of those are useful, some of those are dangerous.
20:56
SPEAKER_02: That notion of discernment is what I'm struggling with how to teach.
21:01
SPEAKER_02: And I wonder how you think about that.
21:04
SPEAKER_02: How do we preserve some sense of, okay,
21:09
SPEAKER_02: here are the things I have to do to advance as a human being,
21:13
SPEAKER_02: to contribute to society that AI can't do for me?
21:17
SPEAKER_01: Well, I'm just thinking about how as someone who writes software for a living,
21:22
SPEAKER_01: if I want this thing to like really do what I'm asking it to do,
21:25
SPEAKER_01: then I have to be like reading that and understanding what it's attempting to do.
21:31
SPEAKER_01: And before I ever get it started, I have to have a very clear idea of the outcome that I want.
21:37
SPEAKER_01: Yeah.
21:59
SPEAKER_01: The amount of text that these can generate is so overwhelming.
22:04
SPEAKER_01: And if you just let it go, you will be washed under by their output.
22:09
SPEAKER_02: Yeah, and tempted to just accept it for what it is because it is so overwhelming.
22:14
SPEAKER_02: But you will wind up with something that is far from what you wanted.
22:17
SPEAKER_01: And maybe the technology will change that and it will get very good
22:20
SPEAKER_01: at discerning what we want from very minimal input.
22:23
SPEAKER_01: And I don't know that that would be desirable in terms of our humanity.
22:28
SPEAKER_01: It's certainly, you know, it's the goal.
22:31
SPEAKER_01: But along the way,
22:32
SPEAKER_01: at every point,
22:33
SPEAKER_01: no matter what,
22:34
SPEAKER_01: if you have something that's just going to do its best
22:37
SPEAKER_01: to provide you whatever you want,
22:39
SPEAKER_01: you better know what you're going to ask for and how to recognize it
22:43
SPEAKER_01: when you see it.
22:44
SPEAKER_02: I think that speaks to one of the asymmetries as well that AI
22:48
SPEAKER_02: at this point seems best suited to folks that sort of understand how to deal
22:52
SPEAKER_02: with its output.
22:54
SPEAKER_02: And that's one thing that I find concerning
22:56
SPEAKER_02: for young people is they don't have this body of knowledge experience and
23:01
SPEAKER_02: the opportunities for apprenticeship appear to be the sorts of jobs
23:06
SPEAKER_02: that AI may be disrupting the most.
23:09
SPEAKER_02: Chrissy, we've opened a lot of doors here.
23:11
SPEAKER_02: You want to touch on any of these in our closing moments?
23:14
SPEAKER_00: I have been thinking a lot about how do we learn discernment if you don't get
23:19
SPEAKER_00: to learn it on the job.
23:21
SPEAKER_00: Actually,
23:22
SPEAKER_00: I think Hank Green has some funny videos about where he's like,
23:25
SPEAKER_00: I got paid to do a job badly at first.
23:30
SPEAKER_00: We all probably did, right?
23:32
SPEAKER_00: Somebody paid us a low wage, we did a job badly, and then we learned how to do it better.
23:36
SPEAKER_00: So now no one's going to pay you to do a job badly anymore because AI can do that.
23:42
SPEAKER_00: It can even probably do it pretty good at an entry level situation.
23:47
SPEAKER_00: And so that gap is real and we're going to have to grapple with that as a society.
23:56
SPEAKER_00: Kind of in the like individual level,
23:59
SPEAKER_00: I am interested in like,
24:01
SPEAKER_00: okay, we need to go back to basics learning, outdoor learning.
24:05
SPEAKER_00: We need to do hands-on learning.
24:07
SPEAKER_00: We need to do group learning.
24:08
SPEAKER_00: We need to like do projects from start to finish that have nothing to do
24:14
SPEAKER_00: with anything on a screen, you know, to really get those projects.
24:19
SPEAKER_00: Kind of really basic like brain neural connections,
24:23
SPEAKER_00: problem solving connections going that can then apply
24:26
SPEAKER_00: to different professional situations and settings.
24:30
SPEAKER_00: And I know that can be unpopular because I think there's a general tendency
24:34
SPEAKER_00: to be like, okay, well, let's learn how to use the AI in the best possible way.
24:38
SPEAKER_00: I think that's actually pretty easy.
24:40
SPEAKER_00: That's the easy part.
24:41
SPEAKER_00: Is this discernment that you're talking about?
24:43
SPEAKER_00: That's the hard part.
24:43
SPEAKER_00: And I think...
24:45
SPEAKER_00: And starting with AI as a tool for learning discernment will be the opposite of
24:50
SPEAKER_00: the most effective way to learn that.
24:52
SPEAKER_02: It's going to say use me all the time in every possible instance and never get off
24:57
SPEAKER_02: the platform.
24:57
SPEAKER_00: Yeah.
24:58
SPEAKER_00: And it just honestly,
24:59
SPEAKER_00: even for me, sometimes I'll run, I'll have it like run a report on something.
25:02
SPEAKER_00: And when I first,
25:03
SPEAKER_00: even with all of my experience,
25:05
SPEAKER_00: even if it's something about,
25:06
SPEAKER_00: if it's about something I know, my first glance, I'm like, oh, this is good.
25:09
SPEAKER_00: This all makes sense.
25:10
SPEAKER_00: Yeah, this is great.
25:11
SPEAKER_00: And then it's only later when I start working with it that I'm like, oh, wait,
25:14
SPEAKER_00: this doesn't actually make sense or this is completely off or this is just
25:18
SPEAKER_00: a hallucination.
25:19
SPEAKER_00: And so if I'm even like fooled at first glance by because it sounds so good,
25:25
SPEAKER_00: the sentences are so well put together.
25:28
SPEAKER_00: So I think asking young people to start at that,
25:30
SPEAKER_00: trying to build their discernment with that level of convincing text,
25:35
SPEAKER_00: that's too hard.
25:36
SPEAKER_00: Don't start there.
25:37
SPEAKER_00: Let's start with, you know, Doing things in the real world with real people and go from there.
25:44
SPEAKER_02: Let's bring it back to AI Realists in our remaining moments here.
25:48
SPEAKER_02: You're producing events, you're engaging in the community, you're doing lots of stuff.
25:51
SPEAKER_02: What are you most excited about in the coming months and how can people get involved?
25:55
SPEAKER_01: I think the thing that's been most exciting to me is watching the bulb go on for people.
26:01
SPEAKER_01: I don't have to be...
26:03
SPEAKER_01: Excellent.
26:26
SPEAKER_00: Yeah.
26:27
SPEAKER_00: Yeah.
26:27
SPEAKER_00: We would love for people to find us at AI Realists, as in the plural, realists.org.
26:36
SPEAKER_00: You can even suggest a new name for us.
26:37
SPEAKER_00: We're open.
26:39
SPEAKER_00: And we are at a really, I think, fun and exciting stage because we are still open to new ideas.
26:50
SPEAKER_00: We're still...
26:51
SPEAKER_00: We're not planning to do all of the ideas ourselves.
26:54
SPEAKER_00: We're pushing things back out to other people.
26:56
SPEAKER_00: It's a chance to get involved.
26:57
SPEAKER_00: It's a chance to engage.
26:59
SPEAKER_00: And honestly, it's a lot more interesting and fun and hopeful than like building a bunker.
27:05
SPEAKER_00: So, you know, come to AI.
27:06
SPEAKER_00: If you're feeling despair, come to AIrealist.org and, you know, we'll have coffee.
27:10
SPEAKER_00: We'll brainstorm.
27:12
SPEAKER_00: We'll do stuff.
27:13
SPEAKER_02: Thanks to you both for kind of going along on this meandering ride today.
27:18
SPEAKER_02: And just thank you for the work you're doing to try to push back against that sense
27:24
SPEAKER_02: of inevitability and helplessness and try to actually do something
27:29
SPEAKER_02: as Rasa implored you to do.
27:32
SPEAKER_02: Yeah, best of luck and we hope to revisit down the road.
27:38
SPEAKER_02: Thanks for listening to A New Angle.
27:40
SPEAKER_02: If you like what we do and want to support us directly,
27:42
SPEAKER_02: please consider a subscription at anewanglepodcast.com.
27:49
SPEAKER_02: A New Angle is recorded at Studio 49, a generous gift from UM alums Michelle and Lauren Hanson.
27:55
SPEAKER_02: Our presenting sponsor is the Headwaters Foundation,
27:58
SPEAKER_02: working side-by-side with Western Montanans to improve the health
28:01
SPEAKER_02: of our communities.
28:04
SPEAKER_02: We are also supported by First Security Bank and the University of Montana College
28:08
SPEAKER_02: of Business, with additional help from CED, Drum Coffee, and Montana Public Radio.
28:15
SPEAKER_02: Keely Larson is our producer.
28:17
SPEAKER_02: John Brownell, Jeff Ament, and John Wicks made our music.
28:22
SPEAKER_02: Thanks a lot, and we'll see you next time.