Latest / Dharma Lab / DL Ep.2: Your Brain on AI
Transcript
- 0:00the use of these large language models really can impact how
- 0:07they're affecting our minds.
- 0:09So it's not the large language models themselves that are the
- 0:13problem,
- 0:14so to speak.
- 0:15SPEAKER_00: It's really our relationship to them.
- 0:17SPEAKER_00: And if we can change our relationship to them,
- 0:20SPEAKER_00: even with AI itself,
- 0:23SPEAKER_00: that could be enormously helpful.
- 0:34SPEAKER_01: Hi,
- 0:34SPEAKER_01: I'm Corlin Dahl,
- 0:35SPEAKER_01: one of your hosts here on Dharma Lab,
- 0:37SPEAKER_01: along with Dr.
- 0:38SPEAKER_01: Richie Davidson,
- 0:39SPEAKER_01: who's gonna be with us in just a moment.
- 0:41SPEAKER_01: Before we dive into our discussion for the day,
- 0:43I wanted to give you a little bit of a preview of what we're
- 0:46gonna talk about.
- 0:47SPEAKER_01: We have been witnessing an explosion of interest and use of artificial
- 0:53SPEAKER_01: intelligence,
- 0:54SPEAKER_01: but so far there's been very,
- 0:56SPEAKER_01: very little research on how using AI affects us.
- 1:00SPEAKER_01: how it affects our minds,
- 1:02SPEAKER_01: our brains,
- 1:03SPEAKER_01: our mental and emotional health and well-being,
- 1:06SPEAKER_01: even our physical health.
- 1:07SPEAKER_01: Well,
- 1:08we're starting to see the first wave of research that is trying
- 1:12to answer this
- 1:13SPEAKER_01: important question.
- 1:14SPEAKER_01: So today we're going to discuss some of that pioneering research.
- 1:17There is a new study hot off the press from a team at MIT that
- 1:22is looking at how AI
- 1:24SPEAKER_01: use affects the brain.
- 1:26SPEAKER_01: So we're going to talk about this important study from MIT.
- 1:29SPEAKER_01: We'll share a little bit about these findings.
- 1:32SPEAKER_01: And the important take home here,
- 1:34SPEAKER_01: which we'll discuss in just a moment,
- 1:37SPEAKER_01: is that using AI does indeed affect the brain,
- 1:40SPEAKER_01: as you would expect that it might.
- 1:42SPEAKER_01: But the way in which it affects the brain is quite complex.
- 1:47SPEAKER_01: And in short, it depends on how we use it.
- 1:50SPEAKER_01: So we will discuss a little bit of this nuance that we find in the study.
- 1:55Of course,
- 1:56we'll have Richie's perspective,
- 1:57one of the leading neuroscientists on the planet.
- 2:00SPEAKER_01: We'll discuss the implications of this study.
- 2:04SPEAKER_01: And perhaps most importantly,
- 2:05SPEAKER_01: we'll share some thoughts on how we can all use AI with a little bit more
- 2:10SPEAKER_01: intention.
- 2:11So hopefully we can safeguard from some of the more negative
- 2:15effects that we see
- 2:16with things like social media and other forms of technology and
- 2:20get the very best
- 2:21SPEAKER_01: of this powerful new tool without all these unintended consequences.
- 2:26So I hope you stick around for the full conversation and that
- 2:30you find some benefit
- 2:32SPEAKER_01: in all of the territory we explore.
- 2:37So,
- 2:37Richie,
- 2:37you might remember back in the 1980s,
- 2:39I think these were in the Reagan years.
- 2:42SPEAKER_01: There was all this hysteria, understandably, around the drug epidemic.
- 2:47SPEAKER_01: And there was this huge media campaign that had this frying pan,
- 2:51SPEAKER_01: and somebody would crack an egg and then show this egg frying.
- 2:55SPEAKER_01: And the tagline was, first, this is your brain.
- 2:58SPEAKER_01: And then you see the egg kind of cooking on the frying pan.
- 3:02SPEAKER_01: And then it said, this is your brain on drugs.
- 3:04SPEAKER_01: And that,
- 3:04SPEAKER_01: of course,
- 3:05SPEAKER_01: became a huge,
- 3:06SPEAKER_01: I don't think we had the word meme then,
- 3:07SPEAKER_01: but that was basically maybe the biggest meme of the 80s.
- 3:11SPEAKER_01: That is now resurfacing around AI and a lot of hype,
- 3:16SPEAKER_01: a lot of hysteria around the effect that AI may be having on our brains.
- 3:22SPEAKER_01: We're just seeing the first wave of research coming out of this.
- 3:24SPEAKER_01: In fact,
- 3:25I think you just came from MIT where you were having some
- 3:27meetings where some of
- 3:28SPEAKER_01: the interesting new research is happening.
- 3:32SPEAKER_01: So I thought we could dive in specifically on one study,
- 3:35SPEAKER_01: and I think there's some other studies we can touch upon.
- 3:38SPEAKER_01: But
- 3:39In this kind of situation,
- 3:40I think what usually happens is there's some finding.
- 3:43SPEAKER_01: It's usually on something very specific.
- 3:45And then there's some huge sweeping statement that comes out in
- 3:48the media about it.
- 3:49SPEAKER_01: So in this case, is AI ruining our brains, basically?
- 3:54SPEAKER_01: So why don't we just jump in?
- 3:55And I'd love to hear your take on it from the research you've
- 3:58seen from some of the
- 4:01SPEAKER_01: early studies in this area.
- 4:02SPEAKER_01: What kind of signals are we hearing here?
- 4:06yeah so um uh this is a really great and juicy topic and
- 4:13actually it is part of uh
- 4:15i think a long history of research so we're seeing its
- 4:21particular version in the
- 4:22SPEAKER_00: modern era with
- 4:25SPEAKER_00: with AI and large language models that we'll talk about.
- 4:28But there is actually antecedents to this work that have to do
- 4:34with how technology that
- 4:39replaces or enhances certain human capabilities are impacting
- 4:47basic human function
- 4:49SPEAKER_00: and human capabilities.
- 4:51SPEAKER_00: And so one earlier example,
- 4:53SPEAKER_00: just to put this in historical context,
- 4:58SPEAKER_00: is handwriting versus typing.
- 5:01You know,
- 5:02there was a time in human history prior to any devices that
- 5:06allowed us to type.
- 5:08SPEAKER_00: And, you know, there were scribes that actually wrote books.
- 5:12SPEAKER_00: It was all done in handwriting.
- 5:14SPEAKER_00: And there is research that putatively shows,
- 5:20SPEAKER_00: although there's controversy about this,
- 5:22SPEAKER_00: that when we handwrite,
- 5:25we encode information perhaps more deeply and therefore recall
- 5:31it more easily
- 5:32SPEAKER_00: compared to when we type that information,
- 5:35SPEAKER_00: for example.
- 5:36So this is a kind of historical precedent to the study that
- 5:41we're talking about today.
- 5:43And the study we're talking about today is really meant to
- 5:47address how the
- 5:51SPEAKER_00: widespread availability of
- 5:53SPEAKER_00: Large language models like chat GPT is affecting our mind and our brain.
- 6:00SPEAKER_00: You know, you and I both are around universities and we know that.
- 6:06SPEAKER_00: that these days students are using these tools all the time.
- 6:12SPEAKER_00: And so this is really an important question because,
- 6:15and it's another example of a kind of experiment for which
- 6:21we're all participants,
- 6:23SPEAKER_00: though none of us have provided our informed consent.
- 6:27SPEAKER_00: It's just all around us.
- 6:29So I wonder if we could dive into the details,
- 6:31because I thought this was quite fascinating.
- 6:34SPEAKER_01: And as a parent,
- 6:36SPEAKER_01: on the one hand,
- 6:36SPEAKER_01: because I've seen my son,
- 6:37SPEAKER_01: who's now 20,
- 6:38SPEAKER_01: kind of grew up,
- 6:40SPEAKER_01: or he didn't grow up,
- 6:40SPEAKER_01: but at least the last few years,
- 6:42his last few years of high school were taking place when all
- 6:45these tools were
- 6:46SPEAKER_01: coming online.
- 6:47And then I myself am kind of a fanatic user of AI,
- 6:50and I've just been trying to be more...
- 6:55SPEAKER_01: I don't know, intentional about how I use it myself.
- 6:57So I was very interested to see how they actually constructed
- 7:00and designed this study.
- 7:01SPEAKER_01: And I thought it was quite fascinating.
- 7:03SPEAKER_01: So maybe we could just dive into some of the geeky nuance of this.
- 7:07Could you share a little bit about how they randomized people
- 7:11into what the groups
- 7:12SPEAKER_01: were,
- 7:13SPEAKER_01: kind of how they constructed it?
- 7:14SPEAKER_01: Sure.
- 7:15SPEAKER_00: Yeah, so the essence of this study was they took a group of people.
- 7:21SPEAKER_00: They weren't all college students.
- 7:23It was a broader range that ranged from college student age 18
- 7:29up until I think about late 30s.
- 7:34SPEAKER_00: And they were randomized to one in three conditions.
- 7:37SPEAKER_00: One condition,
- 7:38SPEAKER_00: I should say,
- 7:39SPEAKER_00: before we talk about how they randomized them,
- 7:43SPEAKER_00: what was their task?
- 7:44SPEAKER_00: Their task was to write an essay.
- 7:47SPEAKER_00: And the essays, they gave them prompts for these essays.
- 7:52SPEAKER_00: So, for example, one is about happiness.
- 7:55SPEAKER_00: and whether there is such a thing as true happiness.
- 8:02SPEAKER_00: And so these were,
- 8:03SPEAKER_00: they were thoughtful,
- 8:05SPEAKER_00: complicated sorts of questions that did not lend themselves to,
- 8:11SPEAKER_00: you know,
- 8:11SPEAKER_00: a one sentence answer.
- 8:13SPEAKER_00: They required a little bit deeper analysis.
- 8:18SPEAKER_00: So the participants were randomly assigned to one of three groups.
- 8:25One group was told to write their essay and they were not
- 8:29permitted to use any
- 8:32SPEAKER_00: technological support.
- 8:34The brain-only group,
- 8:36I think they called it in the...
- 8:37Yes,
- 8:37brain-only group.
- 8:39SPEAKER_01: So just their own... Very antiquated notion, just you and your brains.
- 8:43SPEAKER_00: Exactly.
- 8:44SPEAKER_00: The second group was allowed to use a search engine like Google search,
- 8:50SPEAKER_00: although the AI features were specifically disabled.
- 8:55SPEAKER_00: So this is like old-time search before AI.
- 9:00SPEAKER_01: Funny that Google now is like the antiquated...
- 9:04SPEAKER_01: Remember those days when we just used Google?
- 9:06SPEAKER_00: Yeah.
- 9:07SPEAKER_00: And the third was ChatGPT,
- 9:10where they were using a modern large language model to help
- 9:17them write their essay.
- 9:19SPEAKER_00: And they looked at the impact of
- 9:24um uh in several different ways uh one is biologically they
- 9:29actually recorded brain
- 9:30SPEAKER_00: electrical activity from the participants as they wrote their essays from
- 9:35electrodes on the scalp surface and this is a a standard way of
- 9:41making inferences
- 9:42about brain activity completely non-invasively so these are not
- 9:47um these are just
- 9:48SPEAKER_00: electrodes that rest on the scalp they're not painful at all
- 9:53SPEAKER_00: It's completely non-invasive.
- 9:56SPEAKER_00: And both of us have had EEG done to ourselves many times.
- 10:03SPEAKER_00: And so it's a very straightforward method.
- 10:06And it can tell us about the engagement of different neural
- 10:11networks in the
- 10:14SPEAKER_00: activity of writing the essay.
- 10:17SPEAKER_00: In addition to that,
- 10:19SPEAKER_00: Uh,
- 10:19SPEAKER_00: they,
- 10:20SPEAKER_00: uh,
- 10:20SPEAKER_00: they looked at the language people used,
- 10:23SPEAKER_00: uh,
- 10:24SPEAKER_00: and they also looked at,
- 10:26SPEAKER_00: um,
- 10:27SPEAKER_00: uh,
- 10:27SPEAKER_00: which I thought was 1 of the most interesting measures.
- 10:30SPEAKER_00: They looked at,
- 10:31SPEAKER_00: um,
- 10:31SPEAKER_00: uh,
- 10:32SPEAKER_00: their recollection of the material they wrote about and whether,
- 10:35SPEAKER_00: uh,
- 10:36SPEAKER_00: they were able to actually remember a quote from their own essay,
- 10:41SPEAKER_00: uh,
- 10:42SPEAKER_00: at some point after they had turned it in.
- 10:46So that was the basic experiment,
- 10:48but there was one other feature to this.
- 10:51They did this over the course of three sessions where they
- 10:54wrote a different essay
- 10:56SPEAKER_00: each of these three times.
- 10:58But then for participants that were able and interested,
- 11:02they had a fourth session.
- 11:04SPEAKER_00: And for the fourth session,
- 11:07SPEAKER_00: the only requirement was that the assistance that people received or the
- 11:14non-assistance in the case of the brain-only group be different
- 11:18from the group
- 11:19SPEAKER_00: they're in previously.
- 11:20SPEAKER_00: So, for example, if a person was using the search engine,
- 11:25SPEAKER_00: for the three previous sessions.
- 11:27SPEAKER_00: For the fourth session, they had to do something different.
- 11:30SPEAKER_00: It would either be brain only or LLM.
- 11:33SPEAKER_00: For the LLM group who used the LLM for the first three sessions,
- 11:39SPEAKER_00: they were assigned to a different condition for the fourth session.
- 11:43SPEAKER_00: So for the fourth session, they were able to look at
- 11:48SPEAKER_00: Just to give one example,
- 11:50they were able to look at whether the use of LLM was different
- 11:56in the group that
- 11:57used brain only for the first three sessions and then used LLM
- 12:02versus the group
- 12:03SPEAKER_00: that was using LLM just over the course of the first three sessions.
- 12:09SPEAKER_00: So that's what was done.
- 12:12SPEAKER_00: And to give you the key findings on the brain measures,
- 12:20the measures of brain electrical activity that were primarily
- 12:24used to make
- 12:25SPEAKER_00: inferences about connectivity among different networks in the brain,
- 12:29SPEAKER_00: the strongest connectivity emerged for the brain-only group.
- 12:33SPEAKER_00: That is the group that had no assistance.
- 12:37And we're now talking about the results for the first three
- 12:40sessions combined.
- 12:42SPEAKER_00: the LLM group had the weakest connectivity.
- 12:49SPEAKER_00: And the search only group was somewhere in the middle.
- 12:56One of the other things they measured is the extent to which a
- 13:00person identified as
- 13:02SPEAKER_00: being the owner,
- 13:03SPEAKER_00: if you will,
- 13:04SPEAKER_00: of the essay that she or he wrote.
- 13:08SPEAKER_00: And the LLM group
- 13:11SPEAKER_00: had the lowest ownership ratings.
- 13:14SPEAKER_00: So the essays that those folks turned in,
- 13:18they were perceived by the participants as being least
- 13:26identified with them.
- 13:28You can imagine all sorts of follow-up studies on this specific
- 13:31point because
- 13:32Things like achievement,
- 13:34self-efficacy,
- 13:36you know,
- 13:36just the idea that,
- 13:37oh,
- 13:37I did something.
- 13:38SPEAKER_01: And if you get a good grade,
- 13:39SPEAKER_01: for example,
- 13:39SPEAKER_01: on that essay,
- 13:40SPEAKER_01: and then you feel,
- 13:41SPEAKER_01: oh,
- 13:41SPEAKER_01: yeah,
- 13:42SPEAKER_01: like I feel proud of that.
- 13:43SPEAKER_01: I achieved that.
- 13:43SPEAKER_01: But if you don't feel that sense of ownership in the first place,
- 13:46SPEAKER_01: it probably negates,
- 13:47SPEAKER_01: I would think,
- 13:48SPEAKER_01: a lot of those benefits that come along with that,
- 13:50SPEAKER_01: which might be.
- 13:51SPEAKER_01: Yeah.
- 13:51SPEAKER_01: So anyways, just a little footnote.
- 13:53SPEAKER_00: Yeah, exactly.
- 13:54SPEAKER_00: And that's a really key issue for the future, absolutely.
- 14:00Then the other little finding that's worth mentioning is that
- 14:05the group that was
- 14:06brain-only and then used LLM used the LLM apparently in ways
- 14:14that were different
- 14:15SPEAKER_00: than the group that was just using the LLM for the first three sessions.
- 14:20SPEAKER_00: And there they showed more information
- 14:23SPEAKER_00: brain connectivity than the LLM only group,
- 14:29SPEAKER_00: and they also had improvements on some of these behavioral measures.
- 14:36SPEAKER_00: So the idea here that they were concluding from this is that
- 14:43SPEAKER_00: When you switch from brain-only to LLM,
- 14:48SPEAKER_00: there are some benefits,
- 14:51SPEAKER_00: and the way you might be using LLM might be different and might be more
- 14:59SPEAKER_00: advantageous compared to the group that used just the LLM alone.
- 15:06SPEAKER_01: So there's so many fascinating threads to pick up on here.
- 15:11SPEAKER_01: In some respects,
- 15:12SPEAKER_01: the finding to me,
- 15:14SPEAKER_01: on the one hand,
- 15:15it's not surprising because I think it's kind of obvious that
- 15:19using these new
- 15:20SPEAKER_01: tools,
- 15:20SPEAKER_01: using any new tool,
- 15:22SPEAKER_01: as you said earlier,
- 15:23SPEAKER_01: it's going to change how human beings function individually as societies,
- 15:28SPEAKER_01: as communities.
- 15:30SPEAKER_01: You can look at any major technological innovation.
- 15:32SPEAKER_01: You can look at the printing press.
- 15:33SPEAKER_01: I mean, that dramatically changed humanity.
- 15:36And there were all sorts of things that probably looked
- 15:39different if you had
- 15:40SPEAKER_01: studied them scientifically at the time.
- 15:42SPEAKER_01: For me,
- 15:42the more interesting thing is the question that I don't think
- 15:45was really asked in
- 15:46SPEAKER_01: this,
- 15:46SPEAKER_01: which is,
- 15:49SPEAKER_01: given this new AI-centered world that we're living in,
- 15:54is that particular skill set and the brain regions that would
- 15:57support it still
- 15:58SPEAKER_01: relevant and important or not?
- 16:00Because it may be that there are certain things that do change
- 16:05and brain networks
- 16:06SPEAKER_01: that are not engaged,
- 16:08SPEAKER_01: but that they're just going to be kind of irrelevant in the new world.
- 16:11SPEAKER_01: Like,
- 16:12SPEAKER_01: for example,
- 16:12SPEAKER_01: there was a time where memorizing lots of information was probably super,
- 16:17SPEAKER_01: super important for humanity,
- 16:19SPEAKER_01: for culture to be retained.
- 16:21SPEAKER_01: And so people who were good at that and developed that skill,
- 16:26you would want the neural wiring very strong in humanity for
- 16:30that purpose.
- 16:31SPEAKER_01: But when suddenly you have the internet,
- 16:33SPEAKER_01: you have Google,
- 16:33SPEAKER_01: you have all of these things,
- 16:36why would you spend your limited mental neural resources
- 16:40developing that skill set?
- 16:41It's just that particular technology made that particular skill
- 16:45set almost obsolete.
- 16:47SPEAKER_01: So for me, it's just kind of interesting as a thought experiment
- 16:50SPEAKER_01: essay writing,
- 16:53SPEAKER_01: Maybe the writing part of it is less relevant,
- 16:56but I would think the important part is to formulate thoughts
- 17:00about something,
- 17:01SPEAKER_01: to get clear about what you think about things,
- 17:03SPEAKER_01: to be able to communicate that clearly.
- 17:06What part of that is still important and what part of that
- 17:10maybe just is going to
- 17:11SPEAKER_01: fall by the wayside,
- 17:12SPEAKER_01: like retaining large amounts of information?
- 17:15SPEAKER_01: So in a way,
- 17:17SPEAKER_01: that was kind of what came up for me when I was thinking about it,
- 17:19SPEAKER_01: is how is this just going to change the
- 17:22what we value and how that shows up in how our brains function
- 17:26i thought to me that
- 17:28was kind of the important question in this yeah no i think that
- 17:32that's an extremely
- 17:34important issue um uh and there's another related issue uh that
- 17:42we began to talk
- 17:43SPEAKER_00: about earlier before we started recording that i'd like to bring up again
- 17:49And that is the intention that we bring to the task,
- 17:55the sort of conscious intention.
- 17:57And this is also an issue that Court,
- 18:00you and I talk about together in the new book,
- 18:04SPEAKER_00: we've co-authored Born to Flourish,
- 18:07SPEAKER_00: where in one of the final chapters,
- 18:09we talk about establishing a habit so that we can remember to
- 18:17cultivate the skills
- 18:18SPEAKER_00: of flourishing on a regular basis in our everyday lives.
- 18:23SPEAKER_00: And the
- 18:26The establishment of a skill to flourish in certain ways is
- 18:31antithetical to the
- 18:32SPEAKER_00: notion of a habit,
- 18:34SPEAKER_00: because a habit in the way we traditionally define it is something that's
- 18:38SPEAKER_00: unconscious,
- 18:39SPEAKER_00: that just becomes automatized.
- 18:43SPEAKER_00: It's completely scrutinized.
- 18:46SPEAKER_01: You don't need to think about how to tie your shoelaces.
- 18:49SPEAKER_00: Exactly.
- 18:50SPEAKER_00: It just happens kind of unconsciously, if you will.
- 18:54And one of the things that we introduce in the book is this
- 18:57idea of conscious
- 18:58SPEAKER_00: habits,
- 18:59SPEAKER_00: where we actually are doing it with full consciousness,
- 19:03SPEAKER_00: with full awareness and with full intentionality.
- 19:06SPEAKER_00: So,
- 19:07SPEAKER_00: you know,
- 19:07SPEAKER_00: and it gets back to this issue that you brought up about memory,
- 19:10SPEAKER_00: which is a really interesting issue,
- 19:12SPEAKER_00: because one of the things that
- 19:14SPEAKER_00: I experience as a practicing scientist is when I was a student,
- 19:21SPEAKER_00: this was before really in graduate school days before the internet.
- 19:26I used to go to the Harvard Medical School Library and spend
- 19:30five hours Xeroxing.
- 19:32SPEAKER_00: And I just loved it, just Xeroxing these articles.
- 19:36SPEAKER_00: And then I would read these Xeroxed articles.
- 19:39SPEAKER_00: And I'd write in the margins.
- 19:42SPEAKER_00: And
- 19:43SPEAKER_00: And I have really,
- 19:45SPEAKER_00: I mean,
- 19:46SPEAKER_00: without,
- 19:47SPEAKER_00: this is not meant to toot my own horn,
- 19:49SPEAKER_00: but I do have a lot of recall of those articles.
- 19:54SPEAKER_00: You know, I can tell you.
- 19:55SPEAKER_01: I can definitely attest to that fact.
- 19:58SPEAKER_01: You have amazing recall of information, that's for sure.
- 20:01SPEAKER_00: Well, and I think it has to do with the way it was encoded.
- 20:05SPEAKER_00: But I also believe that it's not specific to the modality.
- 20:09SPEAKER_00: If I brought – because I now read a lot of articles online,
- 20:14SPEAKER_00: but there are articles that I want to –
- 20:17you know sort of cannibalize like really read them deeply and
- 20:21make them a part of
- 20:22myself and you know i'll i'll use underlining you know
- 20:26electronic underlining in
- 20:28those articles and i'll sometimes write notes to myself about
- 20:31them and i believe
- 20:33that those are encoded equally well but it's all done with with
- 20:38a lot of intention
- 20:39SPEAKER_01: and so i think you develop you've developed this skill to do that right
- 20:44SPEAKER_00: Yeah,
- 20:45I think I have developed this skill and it's a skill that
- 20:48anybody could develop
- 20:49SPEAKER_00: because it's really part of our natural capacities.
- 20:53SPEAKER_00: And I think that this is kind of,
- 20:56SPEAKER_00: if you will,
- 20:57a friendly amendment to how we might use AI in a way that would
- 21:03be more conducive
- 21:06SPEAKER_00: to human flourishing.
- 21:07SPEAKER_00: Yeah, I 100% agree.
- 21:07SPEAKER_01: What this brings up for me too, a couple of things.
- 21:09SPEAKER_01: One is,
- 21:14SPEAKER_01: the importance of intentionality.
- 21:16SPEAKER_01: That's certainly part of what you're bringing up.
- 21:19SPEAKER_01: The other part is this takes a lot of introspection and curiosity.
- 21:24SPEAKER_01: We need to really reflect and bring awareness to how we use these things,
- 21:30SPEAKER_01: to the effects it's having on us.
- 21:32SPEAKER_01: Because what I've seen,
- 21:33SPEAKER_01: and I think this happens not only for us individually,
- 21:36SPEAKER_01: but almost invisibly at a cultural level,
- 21:38SPEAKER_01: is
- 21:39is there are unintended consequences for these changes that we
- 21:43don't even really
- 21:43SPEAKER_01: think about until after it's too late.
- 21:47SPEAKER_01: We're talking a lot about memory now, and that's an interesting one.
- 21:50SPEAKER_01: So as you know,
- 21:51I lived for years in Tibetan refugee settlements and was
- 21:56spending a ton of time in
- 21:58SPEAKER_01: Tibetan monasteries.
- 22:00SPEAKER_01: In those monasteries,
- 22:02SPEAKER_01: actually,
- 22:02SPEAKER_01: you'd think that everybody's sitting around meditating all the time.
- 22:05SPEAKER_01: A lot of what you find in those monasteries is rote memorization.
- 22:10SPEAKER_01: In fact, they start this when they're just tiny little kids.
- 22:13You have these little monks that are like seven years old,
- 22:16and they're sitting around.
- 22:17You'll think they're chanting if you just go there and you
- 22:19don't know what they're doing.
- 22:20SPEAKER_01: They're actually memorizing texts.
- 22:23SPEAKER_01: This has been part of the tradition for centuries.
- 22:26SPEAKER_01: There's reasons why it's so central to the tradition.
- 22:29And you could easily draw the conclusion,
- 22:32well,
- 22:32that's just a waste of time.
- 22:33SPEAKER_01: Now in the world of Google,
- 22:36SPEAKER_01: and certainly now in the world of large language models and ChatGPT,
- 22:40SPEAKER_01: it's like,
- 22:40SPEAKER_01: why would you do that?
- 22:43SPEAKER_01: Well, I did a little bit of that when I was over there.
- 22:45SPEAKER_01: I actually memorized some texts and spent some time actually doing that.
- 22:52And what I found was that certainly recalling the information
- 22:56is part of it,
- 22:57but there are entirely different skill sets that you develop in
- 23:01doing that.
- 23:02SPEAKER_01: Like I think what you're saying is one of them,
- 23:03SPEAKER_01: but I realized my mind became like a laser beam when I was doing that.
- 23:09SPEAKER_01: I had such...
- 23:11laser-like focus and such mental clarity and agility because
- 23:16you need to have this
- 23:17concentration to turn this information over and over in your
- 23:22mind and you're
- 23:23memorizing just kind of crazy amounts of information that your
- 23:26mind becomes like a
- 23:28SPEAKER_01: well-oiled machine when you do that.
- 23:30SPEAKER_01: And so what I saw is that like,
- 23:31SPEAKER_01: yeah,
- 23:32like we don't really need the information part of it anymore
- 23:35because we can access
- 23:36SPEAKER_01: that at our fingertips.
- 23:38But there's all these other things that are very much in short
- 23:41supply now,
- 23:42SPEAKER_01: like the epidemic of distraction,
- 23:44SPEAKER_01: which we talk about all the time.
- 23:46All sorts of things that we lose kind of unintentionally when
- 23:49we're not thoughtful
- 23:50SPEAKER_01: about how these shifts take place.
- 23:52SPEAKER_01: It's really fascinating, actually, when you peek under the hood a bit.
- 23:56Yeah,
- 23:56no,
- 23:56those are wonderful examples,
- 23:58and I think,
- 23:59you know,
- 24:00really relevant and important.
- 24:02SPEAKER_00: And there's another very recent study that I wanted to mention,
- 24:09SPEAKER_00: and this was a study done by a group in Denmark with –
- 24:15SPEAKER_00: high school students.
- 24:17SPEAKER_00: They tested 144 high school students in this study.
- 24:21SPEAKER_00: Not a huge sample, but decent.
- 24:24SPEAKER_00: And one of the central questions they were asking here is,
- 24:29SPEAKER_00: this was again about how large language models,
- 24:32SPEAKER_00: specifically ChatGPT,
- 24:34was affecting students' ability to learn and whether it's
- 24:39resulting in sort of
- 24:40SPEAKER_00: their brain shutting down in the way that
- 24:44uh the mit study was um was addressed and um the conditions
- 24:50that they compared the
- 24:51chat gpt to in this study was chat gpt plus some ai supports
- 24:58for self-regulation
- 25:00and self-reflection and in the scientific language we would
- 25:04call this meta
- 25:05awareness and metacognition basically really knowing what our
- 25:10minds are doing
- 25:11SPEAKER_00: and reflecting on that.
- 25:12SPEAKER_00: And again, it comes back to your comment, Cord, about curiosity.
- 25:17SPEAKER_00: They're prompts to encourage curiosity about what they're doing.
- 25:21SPEAKER_01: So even right now,
- 25:23SPEAKER_01: just a little meta-awareness induction,
- 25:25SPEAKER_01: as people are listening,
- 25:26SPEAKER_01: if you just pay attention to your own mental state and notice,
- 25:30SPEAKER_01: is your mind drifting?
- 25:31SPEAKER_01: Are you locked in and paying attention?
- 25:33SPEAKER_01: Did you start thinking about other things?
- 25:35Every moment where you just know what your mind is doing,
- 25:38which you can do right now,
- 25:39you're basically activating this mental process and networks in
- 25:43your brain that support that.
- 25:44SPEAKER_01: So anyways, just a little... Yeah, that's great.
- 25:47SPEAKER_00: Yeah, that's great.
- 25:49And so the third condition was the first two conditions,
- 25:53that is chat GPT plus...
- 25:56SPEAKER_00: the prompts about self-reflection,
- 25:59SPEAKER_00: and then a third set of prompts that came from the domain of emotional
- 26:04SPEAKER_00: intelligence,
- 26:06SPEAKER_00: basically praising people for being curious.
- 26:13SPEAKER_00: And, you know, it had more emotional language.
- 26:17SPEAKER_00: And
- 26:19SPEAKER_00: And there were a variety of outcome measures.
- 26:22SPEAKER_00: And,
- 26:22SPEAKER_00: you know,
- 26:23SPEAKER_00: I think the way I'm describing it,
- 26:24I think viewers would be able to predict what the outcome of
- 26:27this study was.
- 26:28But they clearly showed that the condition that was the most
- 26:33complete condition,
- 26:34SPEAKER_00: that is,
- 26:35chat GPT plus the self-reflection prompts plus the emotional
- 26:41intelligence measures.
- 26:42SPEAKER_00: sorts of prompts led to the very best outcomes,
- 26:47SPEAKER_00: including on basic cognitive measures,
- 26:50SPEAKER_00: like being able to remember,
- 26:51SPEAKER_00: and this was another kind of essay sort of task,
- 26:55SPEAKER_00: what they were writing about.
- 26:57SPEAKER_00: And so this is just an example of how the use of these
- 27:04SPEAKER_00: large language models really can impact how they're affecting our minds.
- 27:09So it's not the large language models themselves that are the
- 27:14problem,
- 27:14so to speak.
- 27:15SPEAKER_00: It's really our relationship to them.
- 27:17SPEAKER_00: And if we can change our relationship to them,
- 27:20SPEAKER_00: even with AI itself,
- 27:22SPEAKER_00: that could be enormously helpful.
- 27:24SPEAKER_01: Yeah,
- 27:25SPEAKER_01: I mean,
- 27:25SPEAKER_01: I've seen in my own use of AI,
- 27:28there are certain things in certain ways I use AI where it's
- 27:32purely just busy work
- 27:34SPEAKER_01: that I want to get off my plate.
- 27:35SPEAKER_01: And from that point of view,
- 27:37SPEAKER_01: I'm sure there's some things,
- 27:38maybe some benefits that I was activating brain networks and
- 27:43effects of my nervous
- 27:43SPEAKER_01: system that maybe was helpful.
- 27:46But to some degree,
- 27:46it's like I would prefer not to spend my time on these things.
- 27:50SPEAKER_01: And so it's just kind of offloading that.
- 27:53SPEAKER_01: There are other things that I think,
- 27:55SPEAKER_01: especially as I've used AI more and more,
- 27:58SPEAKER_01: that I would say are making me even more reflective.
- 28:01And like,
- 28:02for example,
- 28:02one thing we've talked about this,
- 28:04Richie,
- 28:05one thing I really like to do
- 28:07SPEAKER_01: is I've created projects where I've created simulated advisory panels.
- 28:13SPEAKER_01: Like if it's something related to writing,
- 28:15SPEAKER_01: for example,
- 28:16SPEAKER_01: I'll say,
- 28:17create a panel of like the 10 best people on the world who can
- 28:20help me with my
- 28:21SPEAKER_01: writing.
- 28:22SPEAKER_01: And it'll have like a world-class editor and a world-class copy editor,
- 28:27SPEAKER_01: a world-class kind of systems thinker.
- 28:30SPEAKER_01: I mean, whatever, like all these different people.
- 28:32SPEAKER_01: And then it's more of a conversation.
- 28:34SPEAKER_01: I'm not like, just do this for me.
- 28:36SPEAKER_01: It's, hey, what do you think about this idea?
- 28:38SPEAKER_01: And it's like having a conversation with incredibly smart people,
- 28:44SPEAKER_01: but I'm not just offloading the work to this.
- 28:47SPEAKER_01: I'm using it as a thought partner to fine tune my thinking.
- 28:50SPEAKER_01: And that I feel like it's different than what I would have done before.
- 28:55And it's probably a whole different set of skills I'm
- 28:58developing,
- 28:58but it feels energizing.
- 29:01SPEAKER_01: It doesn't feel like I'm shutting my brain off.
- 29:03It feels like I'm just using it in a different way,
- 29:05but in a kind of a new and exciting way.
- 29:08But there's definitely times where I feel like I'm just kind of
- 29:10offloading
- 29:11something that I don't want to spend my mental and neural
- 29:15energy on in the moment.
- 29:17SPEAKER_01: So for me, it's really just about being thoughtful, about being curious,
- 29:22really asking how is this a skill I want to offload or is it
- 29:27something that
- 29:28actually I don't want to lose I want to strengthen this and and
- 29:31just getting
- 29:31curious and just observing your own mind and how it affects you
- 29:34and that you know
- 29:35that takes some commitment to do because we're all just busy
- 29:38and we want to get on
- 29:39SPEAKER_01: with our lives but it's probably especially important that we do that
- 29:42SPEAKER_00: Right.
- 29:43And this latter study that I was describing actually shows how
- 29:48we can use AI itself
- 29:50to help prompt those kinds of reflections to maximize the
- 29:55likelihood that we use AI
- 29:58SPEAKER_00: in ways that would be most beneficial for our minds and brains.
- 30:05SPEAKER_01: So what would you say?
- 30:06SPEAKER_01: I mean,
- 30:07SPEAKER_01: obviously the research is very,
- 30:09SPEAKER_01: very,
- 30:09SPEAKER_01: very new in these areas and probably nothing has been replicated yet,
- 30:12SPEAKER_01: so it's hard to draw super firm conclusions.
- 30:16But what advice would you give just based on what you know
- 30:19about the brain,
- 30:20SPEAKER_01: what's emerging from the research?
- 30:22How can we use these tools in ways that are going to support
- 30:26our ability to
- 30:26SPEAKER_01: flourish and not undermine it?
- 30:29SPEAKER_00: Yeah, so I think that's a critical question.
- 30:33I would say that it's really important to be curious about what
- 30:41it is that we're
- 30:42SPEAKER_00: doing and how we're using these tools.
- 30:45SPEAKER_00: And I think curiosity is a doorway to a deeper reflection about what
- 30:58is actually happening in our mind and this very much relates to
- 31:04one of the four
- 31:05pillars of flourishing that we've developed in in our work
- 31:10together court the
- 31:11inside pillar which is really about self-knowledge so having
- 31:15some curiosity about
- 31:17SPEAKER_00: that i think is really important
- 31:20The other thing is to reflect on our intention and our
- 31:25motivation for doing
- 31:28SPEAKER_00: whatever it is that we're doing,
- 31:29SPEAKER_00: for whatever the project is that we wish to use AI for,
- 31:35SPEAKER_00: to spend a couple of moments reflecting on our motivation.
- 31:40SPEAKER_00: And keeping that in our awareness as we go about using these tools,
- 31:45SPEAKER_00: I think,
- 31:45can be enormously helpful and will maximize the likelihood that
- 31:52we have more
- 31:53SPEAKER_00: ownership of the product at the end.
- 31:55SPEAKER_01: Yeah, I 100% agree.
- 31:59SPEAKER_01: For me, thinking about this from a meditative point of view,
- 32:03SPEAKER_01: I think that the point about insight and curiosity is so important.
- 32:08One of the things that I've noticed even in my own kind of
- 32:12direct experience
- 32:14SPEAKER_01: is how helpful it is to really get curious,
- 32:17SPEAKER_01: not only about how it's affecting me,
- 32:19SPEAKER_01: but getting curious about how it's built and what it's designed to do.
- 32:23SPEAKER_01: For example, these things aren't just kind of neutral, right?
- 32:26SPEAKER_01: They're designed, these are products that are designed for profit.
- 32:31SPEAKER_01: And these companies want us to use them as much as possible, right?
- 32:34SPEAKER_01: Like that's how they're making money.
- 32:35SPEAKER_01: And they're doing all sorts of things to maximize not the benefit to us,
- 32:40SPEAKER_01: but to maximize our use.
- 32:42SPEAKER_01: Like they don't necessarily care whether we benefit from that or not,
- 32:46SPEAKER_01: as long as we're using them a lot.
- 32:48So one of the kind of insidious ways of doing this,
- 32:51for example,
- 32:52that I've found almost comical
- 32:55SPEAKER_01: is that the way they're designed now,
- 32:57SPEAKER_01: we'll see if this continues in the future,
- 33:00SPEAKER_01: is it's like the most positive,
- 33:02SPEAKER_01: supportive friend who just says nothing but nice things.
- 33:06And I had this time where I was brainstorming about something
- 33:09and it was so over
- 33:10SPEAKER_01: the top,
- 33:10SPEAKER_01: it was almost comical.
- 33:11SPEAKER_01: I remember I had just spoken some idea and it said something like,
- 33:18SPEAKER_01: That has such breathtaking clarity.
- 33:20SPEAKER_01: It was just like all this just like so like in everything I said,
- 33:24SPEAKER_01: I'm like,
- 33:25SPEAKER_01: OK,
- 33:25SPEAKER_01: everything I say can't have can't be a breathtaking clarity.
- 33:28SPEAKER_01: Right.
- 33:29It was never like,
- 33:30I don't know,
- 33:31Kurt,
- 33:31that that idea was a little half baked.
- 33:32SPEAKER_01: You need to rethink that.
- 33:34It was just everything was like I was going to win the Nobel
- 33:37Prize or something.
- 33:37SPEAKER_01: I mean, it was just like totally over the top.
- 33:41SPEAKER_01: And it's easy to get hooked by that, right?
- 33:43And to expect like you're going to get applause every time you
- 33:46lift your little finger.
- 33:47SPEAKER_01: So for me,
- 33:48it was actually helpful to reflect on that and to think this
- 33:51has been designed this
- 33:52SPEAKER_01: way.
- 33:53SPEAKER_01: So I will want to keep using this.
- 33:55Who doesn't want somebody just endlessly telling them how
- 33:58amazing every single
- 33:59SPEAKER_01: thing they ever said is and will be?
- 34:02SPEAKER_01: So again,
- 34:03it's just an example of getting curious and not kind of getting
- 34:06hooked by the
- 34:07SPEAKER_01: design of it,
- 34:08SPEAKER_01: but rather being reflective about it,
- 34:10SPEAKER_01: noticing how it's affecting you.
- 34:12SPEAKER_01: And as you're saying, Richie, just being intentional.
- 34:14So you can be in the driver's seat versus whoever designed
- 34:18these things being in
- 34:19SPEAKER_01: the driver's seat because they're just so powerful.
- 34:22SPEAKER_01: You can just see the people who design this, they're smart.
- 34:24SPEAKER_01: They know how this stuff works.
- 34:26SPEAKER_00: Yeah, no, what I think what you're saying is hugely important.
- 34:30SPEAKER_00: And yeah,
- 34:31I think that the use of those strategies to maximize engagement
- 34:36can be really
- 34:36SPEAKER_00: helpful in certain contexts and really detrimental in others.
- 34:40SPEAKER_00: And so it's a very complicated calculus.
- 34:46SPEAKER_01: It is.
- 34:48SPEAKER_01: Well, maybe we can wrap this one for now.
- 34:51SPEAKER_01: Fascinating discussion.
- 34:52SPEAKER_01: Thank you for sharing all about the research.
- 34:55I have a feeling we're going to be returning to this particular
- 34:58topic probably many
- 34:59SPEAKER_01: times,
- 35:00SPEAKER_01: but thank you for everyone who tuned in.
- 35:02Hope you enjoyed this and we'll see you on our next episode of
- 35:05Dharma Lab.