Latest / Azadeh and Guy on AI / All Google All the time!

Transcript

All Google All the time! · Azadeh and Guy on AI

• 0:02
  SPEAKER_01: Welcome to episode 13 of Azadeh and Guy on AI.
• 0:07
  SPEAKER_01: And this time we're all Google all the time.
• 0:09
  And we actually off air,
• 0:11
  we started talking about Nano Banana,
• 0:13
  which is the...
• 0:15
  SPEAKER_01: Picture model, I guess.
• 0:17
  SPEAKER_02: Nano Banana Pro.
• 0:20
  SPEAKER_01: And what I love about that, it's a great product from Google.
• 0:23
  What I love about that is that it is able to edit the photo in
• 0:28
  very precise ways.
• 0:30
  SPEAKER_01: So unlike...
• 0:32
  SPEAKER_01: When I work in ChatGPT,
• 0:33
  SPEAKER_01: and I love producing things with ChatGPT,
• 0:36
  SPEAKER_01: but when I work there,
• 0:37
  SPEAKER_01: the pictures are hard to manipulate.
• 0:40
  SPEAKER_01: Once you make something and you tell it to fix it,
• 0:44
  it either fixes it in ways that you didn't intend or it just
• 0:49
  doesn't fix it.
• 0:49
  It makes another version of it and another version of it,
• 0:52
  but they're all the same.
• 0:54
  SPEAKER_01: And with Nano Banana,
• 0:55
  SPEAKER_01: what I've learned is it's really easy to say,
• 1:00
  just remove this from the background and put this instead or
• 1:04
  just change one piece.
• 1:06
  SPEAKER_01: So you can edit it with commands and it keeps the original text.
• 1:11
  SPEAKER_01: a photo or picture or whatever you want to call that graphic.
• 1:15
  SPEAKER_01: And it keeps working.
• 1:18
  SPEAKER_02: Could you explain more?
• 1:19
  SPEAKER_02: Because I thought maybe for editing photos, we only use prompting.
• 1:27
  I didn't try to edit a photo in NanoVanana,
• 1:30
  but I'm not sure how it's different from ChatGPT.
• 1:35
  SPEAKER_01: It's a lot more precise.
• 1:37
  SPEAKER_01: So when I say... So I created an AI...
• 1:44
  SPEAKER_01: Cowboy, right?
• 1:46
  SPEAKER_01: Podcasting.
• 1:47
  SPEAKER_01: And it put me on a horse.
• 1:50
  SPEAKER_01: It obviously was me.
• 1:52
  SPEAKER_01: And so it put me on a horse.
• 1:53
  SPEAKER_01: And I'm like, no, take the horse away.
• 1:55
  SPEAKER_01: And he just takes the horse away.
• 1:57
  So it's much more precise in the way that it responds to small
• 2:03
  corrections.
• 2:05
  And it keeps the integrity of the original picture that it
• 2:08
  created or photo that
• 2:09
  SPEAKER_01: you upload in fantastic ways.
• 2:12
  SPEAKER_02: So everything was the same.
• 2:13
  SPEAKER_02: It only removed the horse.
• 2:15
  SPEAKER_01: Yes.
• 2:15
  SPEAKER_02: Whereas when you were working with ChatGPT,
• 2:18
  SPEAKER_02: it created a totally new...
• 2:21
  It's either that or sometimes I tell it to change and I just
• 2:24
  get the same thing all
• 2:25
  SPEAKER_01: over again.
• 2:26
  SPEAKER_02: Oh, okay.
• 2:27
  SPEAKER_01: And so...
• 2:28
  SPEAKER_01: It's a much better...
• 2:30
  And for somebody like me who always wanted to manipulate photos
• 2:34
  and to play a
• 2:35
  little bit with the visuals and was not able to because I don't
• 2:39
  have the patience
• 2:40
  SPEAKER_01: for learning a complicated piece of software...
• 2:44
  SPEAKER_00: Yeah.
• 2:45
  SPEAKER_01: And this is really nice.
• 2:48
  SPEAKER_01: Just write the commands.
• 2:50
  And also we're celebrating on Google all the time because
• 2:52
  Gemini 3 Pro is out there.
• 2:55
  SPEAKER_01: Gemini 3 Pro.
• 2:57
  And it's a great new model that I've had very little
• 2:59
  interactions with because who's got time?
• 3:02
  SPEAKER_02: Yeah.
• 3:02
  SPEAKER_02: Yeah, exactly.
• 3:04
  OK,
• 3:05
  but we wanted to actually talk about other things inside the
• 3:07
  Googleverse.
• 3:09
  And the first thing we were going to talk about is actually the
• 3:13
  new addition to
• 3:14
  SPEAKER_01: Google Scholar,
• 3:14
  SPEAKER_01: which is Google Scholar Labs.
• 3:16
  SPEAKER_01: It's not that new.
• 3:17
  SPEAKER_01: It's been around for a while.
• 3:19
  SPEAKER_01: But to be honest, it was there, but it wasn't making a lot of noise.
• 3:24
  SPEAKER_01: So I wasn't thinking about trying it out.
• 3:27
  SPEAKER_01: But now we've tried it out.
• 3:29
  SPEAKER_01: Do you want to talk about what you've learned?
• 3:32
  Yeah,
• 3:32
  and I think the interesting thing is the competition among all
• 3:36
  of those companies.
• 3:37
  SPEAKER_02: I mean,
• 3:38
  SPEAKER_02: we have talked about different AI power tools that you can easily use for
• 3:45
  SPEAKER_02: literature review.
• 3:45
  And I was always thinking to myself,
• 3:47
  what is going to happen to Google Scholar?
• 3:50
  SPEAKER_02: So it's been interesting that they finally came up with something.
• 3:55
  SPEAKER_02: And what I really like about that is that you can...
• 3:59
  go beyond writing keyboards and using different operators
• 4:03
  between them to search
• 4:04
  SPEAKER_02: for something,
• 4:06
  you can actually write your research question and see if it
• 4:12
  looks for any studies.
• 4:14
  SPEAKER_01: And so that's the advantage.
• 4:15
  SPEAKER_01: The advantage is really you can use natural language.
• 4:18
  And the second piece is that instead of just the beginning of
• 4:22
  the abstract taken
• 4:24
  SPEAKER_01: from the article,
• 4:25
  SPEAKER_01: it's actually a summary.
• 4:26
  SPEAKER_01: It's kind of like what you get in a Google search right now.
• 4:29
  SPEAKER_01: You get the AI...
• 4:32
  SPEAKER_01: A short version of everything.
• 4:33
  SPEAKER_01: It feels the same, right?
• 4:36
  SPEAKER_01: In some ways.
• 4:37
  So it gives you a better sense of what's going on because some
• 4:40
  people are good at
• 4:40
  SPEAKER_01: writing abstracts and some people aren't.
• 4:42
  SPEAKER_01: And it wasn't giving you the full abstract.
• 4:44
  SPEAKER_01: So this gives highlights.
• 4:46
  SPEAKER_02: Yeah, yeah.
• 4:46
  SPEAKER_02: It gives you highlights.
• 4:47
  SPEAKER_02: Yes, yes.
• 4:48
  SPEAKER_02: That's very true.
• 4:49
  SPEAKER_01: Is it an amazing piece of software?
• 4:51
  SPEAKER_02: I can't.
• 4:52
  SPEAKER_02: Don't like it.
• 4:54
  SPEAKER_02: We are talking about Google and I expected something spectacular,
• 4:57
  SPEAKER_02: to be honest,
• 4:58
  because I one of the things that I'm interested in is that when
• 5:02
  it comes up with a
• 5:04
  SPEAKER_02: paper,
• 5:05
  I just want to simply go through the references of that paper
• 5:09
  and see what happens.
• 5:10
  SPEAKER_02: So I asked the same thing and I said,
• 5:12
  could you bring up and see if my research question is true
• 5:16
  about any of the
• 5:17
  SPEAKER_02: references that,
• 5:18
  SPEAKER_02: for example,
• 5:19
  SPEAKER_02: one of the results is talking about?
• 5:22
  SPEAKER_02: And the response that I got kind of shocked me.
• 5:25
  SPEAKER_02: It said Google Scholar Labs has not been built for such kind of queries.
• 5:30
  SPEAKER_01: So what is very clear is this is Google's foray into this space.
• 5:37
  SPEAKER_01: And to be honest,
• 5:38
  most of the other AIs that are doing this kind of work around
• 5:44
  the lit review,
• 5:45
  the lit search and all of that seem to lean quite heavily on
• 5:48
  Google Scholar.
• 5:49
  SPEAKER_00: Mm-hmm.
• 5:50
  And I think it is because Google Scholar is one of the best
• 5:53
  aggregators we have out there.
• 5:55
  SPEAKER_01: So if you want a comprehensive search,
• 5:58
  SPEAKER_01: you want to look at specific subsets,
• 6:01
  but you really want to go to Google Scholar because it's
• 6:04
  considerably more precise
• 6:08
  SPEAKER_01: in that way.
• 6:10
  SPEAKER_01: But...
• 6:11
  SPEAKER_01: This is not very much.
• 6:14
  SPEAKER_01: Now,
• 6:14
  SPEAKER_01: the way I think about it,
• 6:15
  SPEAKER_01: and you placed it right,
• 6:16
  is this is really interesting because this is their first foray
• 6:19
  into the space.
• 6:21
  And they have to decide if it's worth it for them to become the
• 6:25
  permanent voice or
• 6:26
  are they okay with being Google Scholar and everybody else
• 6:30
  leaning on them?
• 6:31
  SPEAKER_01: Because the truth is,
• 6:32
  they can stop all these other guys immediately from operating
• 6:37
  by blocking Google
• 6:38
  SPEAKER_01: Scholar from being searched by these AI's.
• 6:42
  SPEAKER_02: That's very true.
• 6:43
  SPEAKER_02: That's a good point.
• 6:44
  SPEAKER_02: Yeah.
• 6:44
  SPEAKER_01: So they could shut it all down and say, you cannot use us.
• 6:50
  SPEAKER_01: We're very sorry.
• 6:51
  SPEAKER_01: We have our own product.
• 6:52
  SPEAKER_01: The second thing is, I'm sure that their next product will be better.
• 6:56
  They're trying to see what kind of queries are out there and
• 7:00
  what do people want to do.
• 7:02
  SPEAKER_01: But it's really interesting how behind they are.
• 7:05
  SPEAKER_02: Yes, yes.
• 7:06
  SPEAKER_02: Yeah, very true.
• 7:08
  SPEAKER_01: That being said,
• 7:09
  it the last week has been the week of Google coming out with
• 7:15
  Nano Banana Pro and
• 7:17
  SPEAKER_01: Gemini 3 and a few other products.
• 7:20
  SPEAKER_01: it is very clear that they're very good at catching up.
• 7:23
  SPEAKER_01: And if they decide that's what they're going to pay attention to,
• 7:27
  SPEAKER_01: they can completely change the playing field.
• 7:31
  SPEAKER_01: The only question I have on my mind is, do they care enough to do this?
• 7:36
  SPEAKER_02: Yeah, yeah, exactly.
• 7:38
  SPEAKER_02: And everybody was super excited about Nona Banana.
• 7:41
  SPEAKER_02: I mean,
• 7:41
  SPEAKER_02: if you go through Substack,
• 7:43
  SPEAKER_02: everybody,
• 7:43
  SPEAKER_02: every single person is talking about that and how great is that.
• 7:47
  SPEAKER_02: And by the way, we also have something on Substack.
• 7:50
  SPEAKER_01: Yes.
• 7:50
  SPEAKER_01: Excellent.
• 7:52
  SPEAKER_01: About that?
• 7:53
  SPEAKER_01: Yeah.
• 7:54
  SPEAKER_01: Excellent.
• 7:55
  SPEAKER_01: I love it.
• 7:57
  SPEAKER_01: But the same thing happened with their video application with Neo3.
• 8:02
  And that is in the beginning,
• 8:04
  they were way behind and now they're really able to produce.
• 8:06
  SPEAKER_01: So there are lots of things happening in the Google space.
• 8:09
  SPEAKER_01: And what we know is if they do want to do that,
• 8:12
  SPEAKER_01: they have the resources,
• 8:13
  SPEAKER_01: they have the capacity and they have the compute to make it happen.
• 8:16
  SPEAKER_01: The only question is, do they care?
• 8:18
  SPEAKER_00: Mm-hmm.
• 8:19
  SPEAKER_00: Mm-hmm.
• 8:20
  SPEAKER_01: And we do know they can because they produced Notebook LM, right?
• 8:24
  And they keep improving that product in ways that are really,
• 8:27
  really interesting.
• 8:28
  So bringing this into this space would not be really
• 8:31
  complicated for them.
• 8:32
  SPEAKER_02: Yeah, exactly.
• 8:34
  SPEAKER_01: All right.
• 8:34
  SPEAKER_01: Let's talk about the other Google News.
• 8:36
  SPEAKER_01: Or do you want to...
• 8:38
  The other thing that we want to talk about is one of the
• 8:41
  Google's reports,
• 8:42
  SPEAKER_02: and it's been published just this month in November.
• 8:45
  SPEAKER_02: The title of the report is AI and the Future of Learning.
• 8:51
  SPEAKER_02: So it was an interesting piece for me,
• 8:54
  especially when I looked at all of the statistics that Google
• 8:58
  actually highlighted
• 8:59
  SPEAKER_02: in that report.
• 9:02
  SPEAKER_01: Tell me more.
• 9:03
  What what what hit you out of all of that that you think this
• 9:07
  is this is where it's at?
• 9:09
  SPEAKER_02: So with education,
• 9:11
  SPEAKER_02: different educational technologies,
• 9:13
  SPEAKER_02: we may feel that we've made a lot of progress.
• 9:15
  But they I think they raised an important point and they said
• 9:21
  learning outcomes are
• 9:22
  SPEAKER_02: sliding globally.
• 9:24
  SPEAKER_02: And they talked about declines in math and reading since 2018.
• 9:31
  SPEAKER_02: And it's interesting because it has started even before the pandemic.
• 9:35
  SPEAKER_02: And they said systems are strained.
• 9:38
  SPEAKER_02: We have funding gaps.
• 9:40
  SPEAKER_02: We have post-pandemic learning loss.
• 9:45
  SPEAKER_02: We always have had teacher shortages.
• 9:48
  SPEAKER_02: And also skill demands are changing fast.
• 9:53
  SPEAKER_01: And so I want to look at it from a few, from a few perspective.
• 9:59
  One is I want to be really careful about pronouncing that
• 10:02
  scores are going,
• 10:03
  SPEAKER_01: that actual reading is going down.
• 10:08
  SPEAKER_01: Just because I want,
• 10:09
  I'm always thinking about what are the measures and what is
• 10:13
  exactly going on.
• 10:14
  SPEAKER_01: Uh-huh.
• 10:15
  SPEAKER_01: And the problem is that we have all kinds of measures and literacy,
• 10:20
  SPEAKER_01: especially literacy in the age of not AI,
• 10:23
  SPEAKER_01: but in the age of the Internet,
• 10:25
  SPEAKER_01: really transformed.
• 10:26
  SPEAKER_01: And we still don't have very good ways to capture it.
• 10:29
  SPEAKER_00: Mm-hmm.
• 10:30
  There were a few efforts in the late 1990s and especially the
• 10:37
  first decade of the
• 10:40
  21st century where we were trying to develop instruments that
• 10:43
  would actually
• 10:45
  capture the set of literacy skills that are necessary in the
• 10:48
  age of the Internet.
• 10:50
  SPEAKER_01: And we know that most of the measures of literacy around the world,
• 10:54
  whether it's the international comparison studies like PISA or
• 10:59
  national efforts,
• 11:01
  they tend to be isolated pieces of reading with a series of
• 11:06
  questions after them,
• 11:07
  SPEAKER_01: which is a way to measure a few things,
• 11:10
  SPEAKER_01: but not everything that you need to be literate.
• 11:12
  SPEAKER_01: All I'm saying is,
• 11:14
  people might be a lot more or a lot less literate in the age of
• 11:20
  the Internet in
• 11:22
  SPEAKER_01: ways that we do not fully understand.
• 11:25
  SPEAKER_01: And so I'm always a little bit worried about these markers and these big
• 11:29
  SPEAKER_01: pronouncements about the stressor and all of that.
• 11:31
  SPEAKER_01: I think something is going on.
• 11:32
  SPEAKER_01: I don't think that.
• 11:34
  SPEAKER_01: And it's not Google's fault.
• 11:35
  SPEAKER_02: No.
• 11:36
  SPEAKER_01: Well, everything is Google's fault.
• 11:37
  SPEAKER_02: Mm-hmm.
• 11:38
  SPEAKER_01: And the reason I say that, I mean, I'm half joking, but half not joking.
• 11:41
  SPEAKER_01: They have more access to what people are doing online than anybody else.
• 11:45
  SPEAKER_02: Yeah, sure.
• 11:46
  SPEAKER_01: What people are looking for,
• 11:47
  SPEAKER_01: how much are they reading,
• 11:48
  SPEAKER_01: how long are they reading,
• 11:50
  SPEAKER_01: what are they doing,
• 11:51
  SPEAKER_01: what is their next query after they read,
• 11:53
  SPEAKER_01: all of that.
• 11:53
  They could actually tell us a lot about how people are behaving
• 11:57
  and how people are
• 11:58
  SPEAKER_01: comprehending online,
• 11:59
  SPEAKER_01: but they haven't done it yet.
• 12:03
  SPEAKER_01: It's just not a priority for them,
• 12:04
  but they actually could tell us a lot about their literacy
• 12:07
  behavior for people who
• 12:08
  SPEAKER_01: are online.
• 12:10
  SPEAKER_01: So that for me is just,
• 12:12
  SPEAKER_01: you could actually help us understand this,
• 12:13
  SPEAKER_01: but now I'm done complaining.
• 12:16
  SPEAKER_01: I'm never done complaining.
• 12:17
  SPEAKER_00: Yeah.
• 12:18
  SPEAKER_01: But the other thing is,
• 12:21
  SPEAKER_01: I'm wondering,
• 12:22
  SPEAKER_01: they talk a lot about personalizing teaching and tutoring.
• 12:25
  SPEAKER_00: Mm-hmm.
• 12:27
  And the data we're getting and the data that a lot of other
• 12:30
  people are getting are
• 12:32
  SPEAKER_01: really interesting around that.
• 12:33
  SPEAKER_01: That is, a lot of people are using AI to teach them things.
• 12:37
  SPEAKER_02: Yeah.
• 12:37
  SPEAKER_02: Mm-hmm.
• 12:38
  SPEAKER_01: Right?
• 12:38
  SPEAKER_01: Explain to me, you know, atomic fusion as if I was a fourth grader.
• 12:43
  SPEAKER_01: Yeah.
• 12:44
  SPEAKER_00: And those kind of things.
• 12:45
  SPEAKER_00: Mm-hmm.
• 12:46
  SPEAKER_01: I am a little bit worried about trying to tailor learning in that way.
• 12:52
  SPEAKER_01: I know everybody is all about that.
• 12:54
  SPEAKER_02: In both ways.
• 12:55
  SPEAKER_02: And what do you exactly mean?
• 12:56
  SPEAKER_01: Personalizing education, personalizing teaching.
• 13:00
  SPEAKER_01: Okay.
• 13:01
  Because the idea is often that the machine will know what level
• 13:05
  you're at and then
• 13:06
  SPEAKER_01: it'll serve you,
• 13:07
  SPEAKER_01: it'll try to push you to do a little bit better,
• 13:10
  SPEAKER_00: whatever.
• 13:10
  SPEAKER_00: And...
• 13:13
  I'm wondering whether that's actually likely to work because I
• 13:17
  can see lots of problems.
• 13:19
  SPEAKER_01: For example,
• 13:19
  SPEAKER_01: if kids figure out that if they make a few errors every time,
• 13:23
  SPEAKER_01: then their life gets easy because it'll keep matching them to that level.
• 13:29
  SPEAKER_01: Kids can game the system rather quickly.
• 13:32
  SPEAKER_01: And so people can figure out, you know, I just keep you.
• 13:38
  SPEAKER_01: And kids do that in real life too, right?
• 13:40
  SPEAKER_01: They tell the teacher they can't.
• 13:42
  SPEAKER_01: There's a whole set of behaviors that can get you off the hook.
• 13:45
  SPEAKER_01: And I'm afraid that with a machine,
• 13:47
  SPEAKER_01: it can actually get worse,
• 13:48
  SPEAKER_01: not better,
• 13:49
  SPEAKER_01: because the machine will keep adjusting.
• 13:51
  And if you make just enough errors,
• 13:53
  then it won't give you a more complicated text.
• 13:56
  SPEAKER_01: Yeah.
• 13:56
  And so you can make your life easy,
• 13:59
  not face the thing that is hard for you.
• 14:02
  SPEAKER_01: Right.
• 14:02
  SPEAKER_00: Yeah.
• 14:03
  SPEAKER_01: By manipulating the machine back.
• 14:06
  SPEAKER_01: And so that's something that I'm always worried about is like,
• 14:10
  SPEAKER_01: when will the machine know that you are actually gaming it?
• 14:12
  SPEAKER_01: Will it ever know?
• 14:15
  Now,
• 14:15
  somebody could look you in the face and say,
• 14:16
  Guy,
• 14:17
  do teachers always know?
• 14:18
  SPEAKER_01: And I'm like, well, no, not always.
• 14:21
  SPEAKER_01: But they can push occasionally to see where it's going.
• 14:25
  So we have to make sure that that and what I'm saying is really
• 14:28
  it's a thing that
• 14:30
  SPEAKER_01: when we think about a quantitative analysis,
• 14:33
  SPEAKER_01: we often talk about overfitting.
• 14:35
  SPEAKER_00: Mm hmm.
• 14:35
  SPEAKER_01: Right.
• 14:36
  And the problem that I have with personalizing teaching is you
• 14:40
  can overfit to the
• 14:41
  student in a way that's actually not productive because the
• 14:45
  student is an active
• 14:46
  SPEAKER_01: player.
• 14:46
  SPEAKER_01: The student has an agency,
• 14:48
  SPEAKER_01: has agency here,
• 14:49
  and they can manipulate that overfitting to not challenge them
• 14:54
  quite as much as we
• 14:55
  SPEAKER_01: think.
• 14:56
  SPEAKER_02: Are you talking about,
• 14:57
  SPEAKER_02: I mean,
• 14:59
  SPEAKER_02: general AI tools,
• 15:01
  SPEAKER_02: for lack of a better word,
• 15:02
  SPEAKER_02: like ChatGPT or that we have a couple of interactions with?
• 15:06
  Or are you talking about AI tools which have been specifically
• 15:11
  designed for schools?
• 15:12
  SPEAKER_02: Yeah.
• 15:13
  SPEAKER_01: I'm more talking about the second one.
• 15:14
  So I'm more talking about systems because what people are
• 15:17
  talking about right now
• 15:18
  are systems that will guide you through learning goals one
• 15:24
  after the other and will
• 15:26
  SPEAKER_01: keep challenging you and will keep upping the ante.
• 15:29
  SPEAKER_02: So about your point for kids...
• 15:32
  playing with AI tools or AI systems and,
• 15:36
  I don't know,
• 15:36
  feed them with inaccurate data.
• 15:39
  SPEAKER_02: I mean,
• 15:39
  if you are talking about AIs that has been specifically
• 15:42
  designed for school
• 15:44
  SPEAKER_02: systems,
• 15:45
  they can kind of track students' progress and then decide about
• 15:51
  their feedback and
• 15:52
  SPEAKER_02: interactions?
• 15:53
  SPEAKER_01: Yes.
• 15:53
  SPEAKER_01: But what I'm saying is kids control their feedback.
• 15:56
  SPEAKER_01: Kids control their achievement.
• 15:59
  SPEAKER_02: But, okay.
• 16:01
  So kids can on purpose game the system to do better or to do
• 16:05
  worse because they're
• 16:06
  SPEAKER_01: like,
• 16:07
  SPEAKER_01: I don't feel like it today or ever.
• 16:09
  And you start learning that just like you can to a certain
• 16:12
  degree manipulate
• 16:14
  SPEAKER_01: people,
• 16:14
  SPEAKER_01: you can manipulate the algorithms just as well or maybe even better.
• 16:18
  SPEAKER_00: Mm-hmm.
• 16:19
  SPEAKER_01: And so I'm a little bit worried about that.
• 16:22
  SPEAKER_01: I think that we need to think about this very,
• 16:24
  SPEAKER_01: very carefully because people are jumping on this and saying,
• 16:26
  SPEAKER_01: and I'm one of them,
• 16:27
  SPEAKER_01: right?
• 16:28
  SPEAKER_01: Occasionally I say,
• 16:29
  this is a great opening to do this personalized learning that
• 16:34
  fits your reading
• 16:35
  SPEAKER_01: level and your challenge level and everything else.
• 16:40
  SPEAKER_01: And what I'm concerned with is...
• 16:43
  SPEAKER_01: Is it overfitting to the individual?
• 16:46
  SPEAKER_01: Right.
• 16:47
  SPEAKER_01: And whenever people talk about reading,
• 16:51
  SPEAKER_01: they talk about interest,
• 16:53
  SPEAKER_01: for example,
• 16:54
  SPEAKER_01: as a prime motivator for kids.
• 16:56
  And my concern always,
• 16:58
  especially working with younger kids,
• 17:00
  is interest is great.
• 17:02
  SPEAKER_01: Right.
• 17:02
  But actually,
• 17:03
  part of education is to open up new areas of interest you
• 17:06
  didn't even know.
• 17:07
  SPEAKER_01: That is my favorite example is if you'd never heard about ancient Egypt,
• 17:12
  SPEAKER_01: how would you ever be interested in it?
• 17:14
  SPEAKER_01: Something has got to start that process.
• 17:16
  SPEAKER_01: And so overfitting to individuals can become a problem.
• 17:21
  SPEAKER_00: Mm hmm.
• 17:22
  SPEAKER_01: And so I'm not saying it's not possible with AI,
• 17:25
  but I'm saying that's the section when I read that section in
• 17:29
  the AI report,
• 17:30
  SPEAKER_01: I'm like,
• 17:30
  SPEAKER_01: there's potential there,
• 17:32
  but there are a lot of possible pitfalls as you design these
• 17:35
  systems to work with
• 17:36
  SPEAKER_01: kids.
• 17:36
  SPEAKER_00: Okay, yeah.
• 17:38
  But I'm super happy that they are entering now,
• 17:41
  going back to Google being an empire.
• 17:43
  SPEAKER_01: They control,
• 17:45
  SPEAKER_01: we talked about this last time,
• 17:46
  they control a good portion of the LMSs that operate in K-12 in
• 17:52
  the United States
• 17:53
  SPEAKER_01: and around the world through Google Classroom.
• 17:55
  And now they brought LM and all of those,
• 17:58
  Notebook LM and all of those into that space.
• 18:00
  SPEAKER_01: So they have a space with literally hundreds of millions of users
• 18:05
  SPEAKER_02: Sure.
• 18:06
  SPEAKER_01: Around the globe.
• 18:07
  And so if they decide that education is where they want to make
• 18:11
  an impact,
• 18:12
  SPEAKER_01: they can rather quickly.
• 18:14
  SPEAKER_01: So all I'm saying,
• 18:15
  I'm paying attention to this report because they have
• 18:17
  tremendous leverage in the
• 18:19
  SPEAKER_01: education space.
• 18:20
  And more than I'm comfortable with,
• 18:22
  I would like multiple companies to compete for that.
• 18:25
  SPEAKER_01: And we do have OpenAI just releasing an OpenAI 5.1 for teachers.
• 18:31
  We're going to look at this hopefully for next time,
• 18:33
  but we'll see about that.
• 18:35
  SPEAKER_01: I'm making plans.
• 18:37
  SPEAKER_01: I'm not committing.
• 18:39
  SPEAKER_01: Maybe not episode 14, but maybe episode 15 or whenever it is.
• 18:44
  SPEAKER_01: Wait with bated breath.
• 18:46
  SPEAKER_01: We will serve it up.
• 18:49
  SPEAKER_02: Exactly.
• 18:50
  SPEAKER_02: And last but not least,
• 18:52
  Notebook LM has released infographic and also a slide deck as
• 18:57
  two additional
• 18:58
  SPEAKER_02: features to their platform.
• 18:59
  SPEAKER_01: Have you tried it?
• 19:00
  SPEAKER_02: Not yet.
• 19:01
  SPEAKER_01: No.
• 19:01
  SPEAKER_01: I haven't either.
• 19:02
  So we're not going to talk about maybe the same time we talk
• 19:06
  about open AI.
• 19:08
  SPEAKER_01: We already have a plan for next time.
• 19:11
  SPEAKER_01: And so join us in episode 14,
• 19:14
  SPEAKER_01: where we're going to keep talking about the things that matter with AI.