Juno News - August 26, 2023


On C2C： Will AI be the end of critical thinking？


Episode Stats

Length

21 minutes

Words per Minute

178.90642

Word Count

3,802

Sentence Count

153

Misogynist Sentences

7


Summary


Transcript

00:00:00.000 I have not been as fearful of artificial intelligence as some people have, because I don't think
00:00:13.620 that human intelligence has often served as well, but that's a bit of a glib joke to start
00:00:18.080 off what is a serious discussion, which is what AI is doing to discourse and to thought.
00:00:24.260 Now, we haven't talked a lot about AI on this show.
00:00:27.280 I've kind of been waiting for the right angle and the right opportunity, and I should say
00:00:31.560 I've been one of these people that has sort of enjoyed the novelty of it.
00:00:34.900 When ChatGPT has come up and you get the ability to just have a quick conversation with this
00:00:40.760 thing and have it give you some response to a question, and there's a program that I've
00:00:46.320 had some fun with called MidJourney, which will create AI-generated images, and you can
00:00:52.720 give it a whole bunch of prompts.
00:00:54.140 I've had a lot of fun with this one.
00:00:55.340 The one that I did, I won't show you because I wasn't thrilled with it, but I asked for
00:00:58.620 like childhood photos of Fidel Castro pushing young Justin Trudeau on a swing, but the AI
00:01:04.560 was getting Fidel Castro and Justin Trudeau's faces mixed up, which maybe makes it smarter
00:01:08.940 than humans.
00:01:09.520 Who knows?
00:01:10.380 And then I also had some fun this morning, and I asked to get like some photographs of
00:01:14.760 Krista Freeland driving, so maybe we can throw those up.
00:01:19.060 Yeah, there we go.
00:01:20.180 These are the samples it gave me.
00:01:21.500 I thought that speed demon, Krista Freeland, fresh off the heels of getting her ticket
00:01:25.580 for going however many kilometers over.
00:01:27.760 That's her basically road racing down some Alberta highway.
00:01:31.620 I like the one on the bottom right myself, although it looks a little terrifying.
00:01:36.000 That one, it looks to be in Ottawa, though.
00:01:37.780 You can see in the back right there, it looks to be a center block that she's just like leaving
00:01:42.640 in the dust there.
00:01:43.980 The one on the top right is good.
00:01:45.120 It's a little aspirational.
00:01:46.080 She's really flying there so much that she needs the space helmet.
00:01:49.860 She's putting on so many miles and going so fast, she has ascended off the ground.
00:01:54.540 So take from that what you will.
00:01:56.420 But for all the fun that AI offers, and yes, there is some, it also has very serious implications.
00:02:03.620 And those implications we have not really fully explored because despite the fact that this
00:02:08.320 technology has been in development for many, many years, it really seems it's only been
00:02:12.940 in the last year that people have started to grapple with the real world implications
00:02:17.960 of it.
00:02:18.500 And, you know, we see this in academia where universities which have had to focus on detecting
00:02:23.940 plagiarism now have this new problem, which is did students just create something original
00:02:29.380 by entering a few prompts for their essay assignments into chat GPT or whatnot.
00:02:36.180 There's a great piece in C2C Journal by Christopher Snook about this called AI,
00:02:41.280 The Destruction of Thought and the End of the Humanities.
00:02:45.000 He is a lecturer with Dalhousie University and a contributor to C2C Journal.
00:02:49.800 And he joins us now, not an AI-generated version, but the man himself.
00:02:54.480 Good to talk to you, Christopher.
00:02:55.480 Thanks for coming on.
00:02:57.040 Thank you, Andrew.
00:02:57.900 Thank you so much for the invitation.
00:02:59.740 So let's start first off with where your issue is with this.
00:03:04.060 Why are you concerned about AI in the context here?
00:03:07.600 Yeah, I can, I suppose I can answer that in a fairly simple way.
00:03:11.280 As you've already indicated, there's a great deal of joy maybe to be had with playing with
00:03:17.000 sort of AI applications.
00:03:18.800 But at the simplest level, I suppose maybe I could say two things.
00:03:22.940 One would be that AI introduces, I'm a humanities teacher, so AI-generated content introduces
00:03:29.520 into the university and into students' lives very easy possibilities of escaping from a certain
00:03:37.240 kind of reflection that may be essential to their development within the context of the
00:03:41.300 humanities historically.
00:03:43.040 But secondly, I think I have a pretty significant concern that AI is actually indicative in many
00:03:48.320 respects of a much longer trend in humanities education in Canada that has fairly uncritically
00:03:55.820 assimilated new technological developments without reflecting on their consequences for pedagogy
00:04:01.720 and education.
00:04:04.440 That's quite an interesting approach to this.
00:04:06.760 And, you know, one thing that I always recall, even from my own time in university, is that
00:04:11.280 essays were very challenging.
00:04:14.060 I would do better at them now, but they were very challenging because you can't really cheat
00:04:17.900 your way through an essay unless you're actually cheating and plagiarizing and whatnot,
00:04:22.720 because it's not just about knowing the facts.
00:04:25.400 You can't Google the answer to the question when you basically have to show your work
00:04:29.280 and show how you arrived at something.
00:04:31.340 And certainly in an academic context, AI has huge implications for that, because all of
00:04:37.040 a sudden someone else could do the thinking with you.
00:04:39.660 I could just give this machine a bunch of different data points and say, formulate an argument
00:04:45.320 for me.
00:04:45.920 And that's something, I mean, I've talked to professors who have already been complaining
00:04:51.100 about the decline in critical thinking in universities.
00:04:53.440 And now we've added this other tool, which maybe can be used for good, but also can further
00:04:58.920 erode people having to come up with these skills on their own.
00:05:03.900 Yeah.
00:05:04.200 What I've tried to do in the article, maybe if I kind of talk about some of the points
00:05:08.000 in the article, that may be helpful for at least giving people a sense of where my concern
00:05:13.020 lies.
00:05:13.480 So my concern really grew out of two things that I saw in the university last year.
00:05:17.640 So the first was, I mean, a remarkable amount of energy and anxiety around the appearance
00:05:22.200 of things like chat GPT, right?
00:05:25.020 Sort of large language models that can produce texts fairly competently, increasingly competently
00:05:32.080 for students with very, very little to no work on their side.
00:05:36.020 So there's a huge amount of anxiety, as you pointed to, Andrew, earlier in your introduction
00:05:39.920 to this conversation.
00:05:42.840 It's a different, it's not even plagiarism in any recognizable sense.
00:05:46.120 So it's just allowing AI to generate texts from the information it's kind of gathered
00:05:51.980 through its internet, through its chat box on the internet.
00:05:54.840 So there was a huge conversation about this in the university.
00:05:57.880 And what I noted was that primarily that conversation was focused on questions of use.
00:06:02.200 And so I spend some time teaching engineers, though I teach humanities to engineers, and they
00:06:05.860 very patiently kind of tolerate this course.
00:06:09.180 I was going to say, that seems like a very, that seems like a very difficult challenge for
00:06:12.180 you.
00:06:12.380 It's a hard sell, it's a hard sell, but they're very patient.
00:06:14.840 And they tolerate this sort of required course on, effectively on the history of technologies.
00:06:20.080 And one of the key things that I've been thinking about since teaching this course is
00:06:24.440 what Neil Postman simply observes, which is that the introduction of every new technology
00:06:28.980 doesn't simply give a new tool to humans.
00:06:31.620 But he sort of coined the idea, or helped kind of articulate the idea that every technology
00:06:37.220 shifts the world ecologically in much the same way that an ecosystem is changed if a new
00:06:43.340 species is introduced, right?
00:06:44.900 So it's not just that we have all of a sudden AI, but rather that the whole world shifts
00:06:49.280 around the availability of these new technologies.
00:06:52.180 And embedded in these technologies are certain assumptions about what it is to be human.
00:06:56.940 And so it's a bit of a rambling response to your previous comment.
00:07:01.800 But what I noticed in the university over the last six or seven months is that the conversation
00:07:07.040 has been almost exclusively focused on use.
00:07:10.240 There have been some people who are sort of diametrically opposed to the appearance of
00:07:15.880 AI in any form in the university.
00:07:17.780 I kind of tend in that direction, certainly for the humanities.
00:07:21.100 Others who are much more supportive of the use of AI in various ways to facilitate writing.
00:07:26.980 But regardless of where one stood on, it stands on the use of AI, I've noticed that very few
00:07:32.360 people are asking deeper questions, such as what kind of world does AI produce and what
00:07:37.420 kind of worldview or what sort of assumptions are built into the technology?
00:07:42.180 And it's there that I think universities need to really be careful about the implementation
00:07:46.000 of AI, partly because I think AI actually reveals, it's a bit apocalyptic, that is, it kind of
00:07:52.900 reveals something about the nature of higher education in Canada that's been developing
00:07:56.980 for years.
00:07:57.740 And we could talk about that sort of narrowing a viewpoint diversity to different aspects
00:08:01.860 of the university life.
00:08:03.560 But also, I think the other thing that I think has been missed is that AI, AI generated texts
00:08:11.000 pushes against certain proclaimed positions, moral positions that the university has adopted
00:08:17.720 in the last, maybe in the last decade.
00:08:19.620 One aspect that this springs to mind, and you address it in the piece, in one section anyway,
00:08:25.640 notably, is the idea of bias.
00:08:27.660 And, you know, facts, in theory, are neutral, and they do not have a political persuasion.
00:08:33.040 It's the assembly of facts and the composition of various facts that you can use to sort of
00:08:38.140 demonstrate something that is a bit more biased.
00:08:40.740 And one thing we've seen in AI is how it's providing, it's doing the thinking for you in theory.
00:08:47.600 But the problem with that, among others, is that it is producing a biased outcome.
00:08:53.660 It's producing a biased response.
00:08:55.520 I mean, I had once, when I was first playing around with ChatGPT, a debate with a machine.
00:09:00.060 So the joke was on me about what a woman is.
00:09:02.540 And it was interesting seeing this machine twist itself into all of these many logical
00:09:06.620 knots about trying to answer this question.
00:09:09.400 But it was actually quite terrifying how it started giving me the talking points I would
00:09:13.980 expect if I were having this with some university diversity administrator.
00:09:17.980 And it started telling me about inclusivity and tolerance and women can come in any forms.
00:09:22.720 And there is something there in which AI is basically telling people that there is one
00:09:29.360 way to construct a thought when it does this, that you aren't actually able to assemble facts
00:09:35.320 into different worldviews.
00:09:37.560 Yeah, I mean, I think that that's true.
00:09:39.900 And certainly the studies have varied and in some cases disagree a little bit with one
00:09:44.140 another about where the biases are found in AI technologies.
00:09:48.180 Though some people have, there's been some studies that have tried to argue that there's,
00:09:51.980 because some of the early scraping of the internet focused primarily on Reddit sites,
00:09:57.760 that there was a kind of conservative or male bias somehow in the technology.
00:10:03.700 But it seems to me fairly clear now that the technology seems to be biased fairly clearly,
00:10:08.480 I think, in the other direction in terms of the kinds of sources that it's recycling when
00:10:15.560 it's when it's producing kind of mashup texts.
00:10:18.120 So I began the article and this is one of the things, maybe a way of thinking about AI in a
00:10:22.200 broader context or sort of moving back from the technology to think about what it has to
00:10:26.180 say about universities.
00:10:27.740 I began the article in C2C really just by reflecting on the fact that
00:10:31.760 a kind of formulaic response to the work of pedagogy has become characteristic of universities
00:10:39.660 generally in Canada.
00:10:41.200 And it's in part indicated through the demand that applicants for university positions
00:10:46.220 complete diversity, equity, and inclusion statements, diversity statements as part
00:10:51.040 of their application packages.
00:10:52.280 And kind of famously, or if you follow these sort of stories infamously, depends on what
00:10:58.500 one thinks about all these things, a professor in the United States asked ChatGPT to produce
00:11:03.840 a diversity document just last year.
00:11:05.860 And he was astounded to see, just as you've described, the speed with which ChatGPT was
00:11:12.060 able to reproduce all of the talking points and all of the assumptions of a fairly kind of
00:11:17.720 middle-of-the-road Canadian higher education position on issues of diversity, inclusion,
00:11:24.340 and equity.
00:11:25.880 To my mind, what that revealed was that we're sort of beginning to operate in the university
00:11:30.140 in a world that is fairly, at the very least, formulaic in its expectations of whatever diversity,
00:11:38.320 inclusion, and equity may be about, or diversity, equity, and inclusion.
00:11:41.920 So it was from there, really, in the article that I wanted to try to see or to explore how
00:11:48.440 is it that the technologies that are available to us now in the university and that have been
00:11:52.740 slowly growing in their implementation in the university over the last 15 or 20 years
00:11:59.760 may actually be both accelerated by the advent of AI, but may also be, in a certain sense,
00:12:08.340 pointing towards AI, that is to say, pointing towards a world in which a kind of formulaic
00:12:13.980 regurgitation of information becomes a kind of normative expectation of students, even
00:12:19.040 in the context of their degrees.
00:12:20.480 So that's sort of where the article began.
00:12:24.120 So just to sort of maybe make a connection with what you found when you poked and prodded
00:12:27.800 ChatGPT, that it tends to kind of produce fairly predictable results relative to certain
00:12:34.100 questions.
00:12:34.720 Well, there is also to this, I mean, the most, I guess, cogent defense I hear of AI is that
00:12:42.860 AI is little more than a mirror to the existing world.
00:12:46.600 I mean, AI is not really formulating its own materials that it's not drawing from the trove
00:12:53.340 of inputs.
00:12:54.400 Now, obviously, individual inputs can be manipulated, and we also have terms of use that govern it.
00:13:00.480 I'm trying to bring us away from the use discussion of this that you were talking about earlier,
00:13:04.560 but I guess in that sense, is this just reflecting an existing problem, or is this making it worse?
00:13:12.980 Right, yeah, that's a very good question.
00:13:14.520 I would say from, there might be two things to say about that.
00:13:17.220 On the one hand, from my perspective, and this was maybe my concern with the conversation so far
00:13:22.900 in higher education in Canada about AI, its preoccupation with questions of use has really prevented
00:13:28.620 people from asking a much slower and more difficult question, which is to say, is this actually a
00:13:34.500 benefit, or is it simply a reflection of the world we're in, or is it making things worse?
00:13:40.640 So I think that that deeper question about the kind of ecosystem consequences or cultural
00:13:46.140 consequences of AI is not really being asked.
00:13:50.780 So that is to say, AI at some level is a kind of metaphor, in much the same way one might think
00:13:57.300 about COVID as a kind of, we can think about it as an illness, but we can also think about
00:14:02.840 COVID response, at least, as a bit of a metaphor of our contemporary cultural moment.
00:14:08.060 So there is that mirroring back.
00:14:10.160 But I would say from the perspective of pedagogy, AI raises some very deep questions that, to
00:14:15.700 my mind, intensify problems that were already present.
00:14:18.620 So it's not so much that it simply introduces a newness that's radical, but intensifies certain
00:14:24.920 very particular problems.
00:14:26.800 So one of those problems, I think, is connected to the use of devices generally for humanities
00:14:34.760 education in particular.
00:14:36.680 So one of the things I think many of us have experienced is the extent to which screens and
00:14:41.880 screen reading and iPhones or cell phones, the extent to which they actually produce in us
00:14:46.500 kind of habits of scanning, a kind of hyper-attention, what one scholar calls forms of hyper-attention,
00:14:51.140 not focused attention or contemplative reflection, but a kind of hyper-attention that actually
00:14:57.100 tends to kind of lead us towards a certain kind of rashness in our decision-making.
00:15:02.260 So that's one deep and profound concern I have, especially when institutions seem to be dominated
00:15:07.380 by certain sets of political commitments that ought themselves to be subject to serious
00:15:12.780 reflection and consideration, right?
00:15:14.460 So if we're in an environment where there's certain assumptions about what political positions
00:15:18.580 are normative, it needs to be the case that those can be thought about deeply and reflectively.
00:15:23.920 And if we're using technologies that limit that capacity, then we're in a little bit of
00:15:27.800 danger.
00:15:28.660 But the other one, I think for me, from the perspective as a teacher, is that, and this
00:15:33.260 would be, I mean, I'm affiliated with the classics department and I spend a lot of time in the last
00:15:38.000 years teaching Augustine, a sort of famous foundational voice for the Western world.
00:15:43.340 And Augustine is one of many thinkers who highlights the fundamental role of memory in the
00:15:49.140 constitution of our personalities, the sort of crucial role that memory plays.
00:15:53.040 And it's kind of essential to technologies like Chachi BT that we offload or offshore the
00:15:59.580 faculty of memory to the technological device, right?
00:16:03.460 It does the work for us, right?
00:16:05.140 So I don't struggle with Augustine or Dante or Homer or any of those things.
00:16:09.680 I let Chachi BT do the struggle in a certain sense.
00:16:12.660 I mean, it's not really struggling, but I let it do the amalgamation of opinion making,
00:16:16.440 information, and I'm left passive in that response.
00:16:20.100 So in that sense, I think the technologies actually inhibit the kind of interior dialogue
00:16:25.320 that's fundamental to education, but that's also fundamental to being a free person in the
00:16:30.300 world, right?
00:16:30.800 Hannah Arendt, I think, points this up very, very powerfully in her reflections on totalitarianism.
00:16:35.740 If we can't have a dialogue with ourselves, if we're pulled out of ourselves endlessly
00:16:40.200 and offshore, even our memory, we lose the ability to actually be free agents in the world.
00:16:47.180 So these are some of the things that I'm very concerned about at the level of pedagogy,
00:16:50.740 which is why I tend to a pretty puritanical, I suppose, relationship to AI when it comes to
00:16:56.440 at least to humanity's classrooms.
00:16:58.040 I recognize AI has different applications in different contexts.
00:17:02.260 It's funny, at the risk of oversimplifying it, I think of, you know, a movie that I've
00:17:06.900 watched that, you know, say is two hours long.
00:17:09.240 I could find out what happens in that movie in about 60 seconds by just reading a plot
00:17:14.340 synopsis on Wikipedia.
00:17:15.760 But I don't do that.
00:17:16.720 I watch the movie because there is something in that process.
00:17:19.800 You feel, you see, you learn, you get insights.
00:17:22.800 It's the same as why, you know, despite the fact that I may not have taken this advice when
00:17:26.120 I was in high school, reading the Coles notes of something is not the same as reading the
00:17:31.420 thing itself.
00:17:32.120 I mean, I could get chat GPT to say, you know, give me some bullet points that I can bring
00:17:36.240 up in tutorial about, you know, the cave or something.
00:17:38.980 But that doesn't mean I've done that.
00:17:40.460 So you're quite right.
00:17:42.100 And I also wonder, I mean, to appeal to your department, the classics, if you were to input
00:17:46.660 into chat GPT, the most beautiful works of literature of classics that you'd ever seen
00:17:52.940 and said, create something like this, could it do that in your view?
00:17:57.240 Could it create the beauty that we have seen from all of these people thousands of years
00:18:02.440 ago?
00:18:03.600 Yeah, that's a very interesting question.
00:18:04.780 That's a very hotly debated topic, as you may know.
00:18:06.980 Of course, in the world of visual arts, someone recently was awarded a prize, right,
00:18:11.220 in the visual arts for an artificially created, produced image, all kinds of, of course, very
00:18:19.400 deep ethical questions around AI and its accumulation of information and how that happens.
00:18:25.520 But, you know, from my perspective, for me, the answer to that question was really given
00:18:30.000 quite beautifully by Nick Cave recently.
00:18:31.760 Nick Cave, the Australian singer-songwriter, was asked this question.
00:18:36.980 A fan sent him a poem that ChatGPT had written when he asked ChatGPT, write me a poem or a
00:18:44.180 song in the style of Nick Cave.
00:18:46.480 And Nick Cave's response was to say that even if it were a good song, which Nick Cave refused
00:18:52.600 to concede that it was a good imitation, he said that the problem is that ChatGPT, artificial
00:18:59.060 intelligence, has been nowhere and suffered nothing.
00:19:02.020 And to be human in the world at all, as someone like Jordan Peterson is constantly reminding
00:19:07.720 us, is to suffer and out of that suffering either to sort of produce meaning in the world
00:19:12.340 and in our lives.
00:19:14.180 And I've been fairly persuaded by Nick Cave that no matter how close the approximation one
00:19:19.780 might be able to artificially reproduce, the fact that the technology has itself been nowhere
00:19:25.440 and suffered nothing means that that material can have very little consequence for me as
00:19:30.680 someone who lives in the world with all of its fragility.
00:19:33.300 So, yeah, so I guess that would be my answer, which isn't, I mean, yeah, maybe not the best
00:19:39.180 answer.
00:19:39.720 No, it is interesting.
00:19:41.040 And now I'm like geeking out on this topic myself.
00:19:43.860 So I think we'll have to have you back on in another show.
00:19:46.160 But, you know, I remember when I did tutorials in various classes in university, the one thing
00:19:52.060 that was always so critical when you were understanding a work was to understand the author and the
00:19:57.300 context in which they wrote a particular work.
00:19:59.940 And even if the author is some professor who's still alive, understanding how that professor
00:20:04.560 came about, you know, you read, for example, a, you know, a dissertation and you say, oh, well,
00:20:09.060 this was an environmental historian.
00:20:10.680 Why were they writing about this issue?
00:20:11.960 And with ChatGPT, that context is eroded because there is no human context or it's an amalgamation
00:20:19.460 of, you know, 150 human contexts that you don't actually know about and can't see.
00:20:23.680 So I think that's a less elegant way of describing what you've shared from Nick Cave there.
00:20:28.660 And I thank you for it.
00:20:30.120 The piece in C2C Journal is AI, the Destruction of Thought and the End of the Humanities by Christopher
00:20:36.340 Snook.
00:20:36.700 And they also have another part of this series written by Gleb Lysik, who we had on the show
00:20:41.080 a couple of weeks ago about something else entirely.
00:20:44.200 Christopher, thanks so much for coming on.
00:20:45.960 Good to talk to a real human in this day and age.
00:20:48.840 Thank you, Andrew, very much.
00:20:49.880 I appreciate it.
00:20:50.600 Thanks for listening to The Andrew Lawton Show.
00:20:53.180 Support the program by donating to True North at www.tnc.news.
00:20:58.640 Thank you, Andrew.