Latest / Azadeh and Guy on AI / From Classrooms to Labs

Transcript

From Classrooms to Labs · Azadeh and Guy on AI

• 0:00
  SPEAKER_00: Hello, everyone.
• 0:01
  SPEAKER_00: We are recording the 10th episode of Azadeh and Guy on AI.
• 0:07
  SPEAKER_01: I'm Guy Trenin,
• 0:08
  and I'm a professor of innovative learning technologies here at
• 0:11
  the University of
• 0:12
  SPEAKER_01: Nebraska-Lincoln.
• 0:14
  SPEAKER_00: And my name is Azadeh.
• 0:15
  SPEAKER_00: I'm a third-year PhD student in innovative learning technologies,
• 0:20
  SPEAKER_00: again,
• 0:20
  SPEAKER_00: at the University of Nebraska-Lincoln.
• 0:23
  SPEAKER_01: And just yesterday,
• 0:25
  SPEAKER_01: actually,
• 0:26
  I was in Kearney for the Nebraska Educational Technology
• 0:29
  Association meeting.
• 0:31
  SPEAKER_01: And this is our fall meeting.
• 0:33
  And I had the opportunity to present about AI,
• 0:36
  but also to see quite a few things.
• 0:39
  SPEAKER_00: Okay.
• 0:39
  SPEAKER_00: Okay.
• 0:40
  We are really interested in hearing what you saw and what you
• 0:44
  discussed and what
• 0:46
  SPEAKER_00: you talked about.
• 0:46
  SPEAKER_00: Okay.
• 0:47
  SPEAKER_01: So my first interaction was actually with Lynn Herr,
• 0:51
  SPEAKER_01: who is a good friend and a graduate of our program.
• 0:54
  SPEAKER_01: And I asked her, how much is there about AI at the conference?
• 0:59
  SPEAKER_01: And she looked at me and said, I think it's all about AI.
• 1:02
  SPEAKER_02: Yeah.
• 1:03
  SPEAKER_01: So other things are happening.
• 1:05
  SPEAKER_01: And the truth is there were a few other things.
• 1:07
  SPEAKER_01: But AI is definitely on everybody's mind.
• 1:09
  SPEAKER_01: And that was very, very clear.
• 1:10
  SPEAKER_01: There were multiple sessions about the ways to use AI.
• 1:14
  SPEAKER_01: And it got really,
• 1:16
  SPEAKER_01: really interesting,
• 1:17
  SPEAKER_01: mostly because I think about conferences in doing two things.
• 1:23
  SPEAKER_01: I think it helps us learn more.
• 1:27
  SPEAKER_00: Sure.
• 1:29
  But the other thing that conferences do is it to help us
• 1:32
  situate ourselves and say,
• 1:34
  SPEAKER_01: is this what everybody's talking about?
• 1:36
  SPEAKER_01: Is there something I'm missing?
• 1:38
  SPEAKER_01: Or are we all asking the same question?
• 1:41
  SPEAKER_01: So I do have a quick answer.
• 1:42
  SPEAKER_01: We're all asking the same question.
• 1:45
  SPEAKER_01: That is, this is so fast and so new that it is worth thinking about
• 1:52
  SPEAKER_01: how do we do this?
• 1:53
  SPEAKER_01: Where do we do this?
• 1:54
  SPEAKER_01: Where does it make sense?
• 1:55
  SPEAKER_01: It's becoming very,
• 1:56
  SPEAKER_01: very clear that everybody's struggling with this,
• 1:58
  SPEAKER_01: whether they're district leaders,
• 2:01
  SPEAKER_01: teachers,
• 2:02
  SPEAKER_01: technology specialists,
• 2:04
  SPEAKER_01: whatever they are,
• 2:04
  SPEAKER_01: they're asking similar questions.
• 2:06
  SPEAKER_00: So let me be more specific, Guy.
• 2:09
  SPEAKER_00: I know that you are doing a lot of research on AI right now.
• 2:12
  SPEAKER_00: Actually, we are doing a lot of them together.
• 2:15
  But did anything stand out to you while you were talking with
• 2:20
  other participants?
• 2:22
  SPEAKER_01: I think that there's what I could gather.
• 2:25
  SPEAKER_01: And again, this is where practitioners go to learn from each other.
• 2:29
  And I think that the thing I heard is nobody knows and
• 2:35
  everybody's thirsty for more
• 2:37
  SPEAKER_01: reliable information.
• 2:38
  SPEAKER_01: So there are a lot of ideas and a lot of things,
• 2:41
  but people are looking for a little bit more certainty because
• 2:44
  their feeling is
• 2:45
  SPEAKER_01: this is moving really fast.
• 2:46
  SPEAKER_01: So,
• 2:46
  SPEAKER_01: for example,
• 2:47
  SPEAKER_01: I was at a session,
• 2:48
  SPEAKER_01: Lynn Herr ran it,
• 2:50
  SPEAKER_01: and she talked about assessment with AI,
• 2:54
  SPEAKER_01: which more specifically regenerative AI and more specifically not exactly
• 2:58
  SPEAKER_01: assessment,
• 2:58
  SPEAKER_01: although we talked about assessment,
• 3:00
  SPEAKER_01: but providing feedback.
• 3:01
  Do you mean creating quizzes when you're talking about
• 3:05
  assessment or other aspect
• 3:06
  SPEAKER_00: of assessment?
• 3:07
  SPEAKER_01: It was really about creating feedback with the thinking around,
• 3:12
  SPEAKER_01: for example,
• 3:13
  how do we help teachers respond to 120 essays that come all in
• 3:21
  the same week,
• 3:22
  SPEAKER_01: right?
• 3:22
  SPEAKER_01: Yeah.
• 3:23
  I'm overwhelmed when I have 22 papers from undergraduate
• 3:26
  students or graduate
• 3:27
  SPEAKER_01: students in my class.
• 3:29
  But if you're an English teacher or if you're teaching foreign
• 3:32
  language or whatever
• 3:33
  SPEAKER_01: it might be,
• 3:35
  SPEAKER_01: especially in high school in the United States,
• 3:37
  SPEAKER_01: you're getting 100 plus assignments,
• 3:39
  SPEAKER_01: sometimes weekly.
• 3:41
  SPEAKER_00: That's overwhelming.
• 3:42
  SPEAKER_01: Yes, it is overwhelming.
• 3:43
  SPEAKER_01: And this is exactly where I can step in.
• 3:45
  SPEAKER_01: So we talked a lot about this.
• 3:47
  SPEAKER_01: But also entering that discussion very early is the idea of transparency.
• 3:53
  SPEAKER_01: And that is you have to tell your students what you're doing.
• 3:56
  And you've got to allow them to think through it and react to
• 4:02
  it and not make them
• 4:03
  SPEAKER_01: surprised that you've used AI.
• 4:05
  SPEAKER_01: If they discover that halfway through,
• 4:07
  just like you as a teacher thought that everybody was writing
• 4:10
  their own essays,
• 4:11
  SPEAKER_01: and then you find out that they haven't,
• 4:14
  SPEAKER_01: you know,
• 4:15
  SPEAKER_01: there's going to be a moment where you check yourself.
• 4:16
  SPEAKER_01: Our students are experiencing very similar things.
• 4:19
  SPEAKER_01: So we talked about the need for transparency.
• 4:23
  SPEAKER_01: So this goes well.
• 4:24
  SPEAKER_01: Right.
• 4:25
  SPEAKER_00: Did you observe any kind of trend?
• 4:28
  SPEAKER_00: I mean,
• 4:30
  I'm asking this question because I was going through the
• 4:32
  responses that we are
• 4:34
  SPEAKER_00: getting from students,
• 4:37
  SPEAKER_00: from teachers,
• 4:37
  SPEAKER_00: sorry.
• 4:38
  We are conducting a survey on culturally responsive teaching
• 4:42
  using generative AI,
• 4:44
  SPEAKER_00: and we talked about it in our last episode.
• 4:47
  SPEAKER_00: So I was going through the responses that we are getting from teachers,
• 4:51
  SPEAKER_00: and one of them said,
• 4:52
  I'm going back to pen and paper system because I want to make
• 4:56
  sure that the
• 4:56
  SPEAKER_00: students are not cheating and using generative AI.
• 5:01
  SPEAKER_00: Yeah.
• 5:03
  SPEAKER_01: There was a little bit of that, but not a lot.
• 5:07
  SPEAKER_01: But that's also because this is a self-selected crowd.
• 5:10
  These are people who came to a conference about technology and
• 5:14
  more than that chose
• 5:15
  SPEAKER_01: to come to a session about AI.
• 5:19
  SPEAKER_00: So they are more receptive about technology and innovation.
• 5:23
  SPEAKER_01: Absolutely.
• 5:24
  SPEAKER_01: But...
• 5:25
  The goal was to really talk through how do we talk to our
• 5:29
  colleagues around these
• 5:30
  SPEAKER_01: issues and,
• 5:31
  SPEAKER_01: of course,
• 5:31
  SPEAKER_01: our students to make sure,
• 5:33
  SPEAKER_01: again,
• 5:33
  SPEAKER_01: it's equitable,
• 5:34
  SPEAKER_01: but also thinking about the upside and not just the downside.
• 5:38
  SPEAKER_01: And I think that those conversations are really what helps.
• 5:41
  SPEAKER_01: If you have the conversations,
• 5:43
  if you play with AI enough to get a sense of what it does well
• 5:47
  and what it doesn't.
• 5:49
  you can start building things that would help you without
• 5:54
  having problems with students.
• 5:55
  SPEAKER_01: But that reaction,
• 5:56
  SPEAKER_01: I'll be honest,
• 5:58
  SPEAKER_01: I mean,
• 5:58
  SPEAKER_01: we're teaching together a class,
• 6:01
  and one of the things I'm asking students to do is to write
• 6:04
  notes in handwriting,
• 6:05
  SPEAKER_01: not really because of AI,
• 6:07
  SPEAKER_01: but because I think that it is important,
• 6:10
  SPEAKER_01: and I'm looking for alternative representations.
• 6:13
  SPEAKER_01: But it's doing the same thing.
• 6:15
  And now if they insist to do it digitally,
• 6:18
  I will let them do it digitally again.
• 6:20
  SPEAKER_01: It's that combination of ways to express ourselves.
• 6:24
  SPEAKER_01: But that's definitely something I saw at the conference.
• 6:27
  The second thing that was really a good conversation is
• 6:31
  everybody is looking for
• 6:32
  SPEAKER_01: good sources and
• 6:34
  on our Substack just this sunday i talked about where do i get
• 6:38
  my information and
• 6:39
  that seemed to have resonated with a lot of people so a lot of
• 6:42
  people actually
• 6:42
  walked up to me and said oh i read that and i'm going to start
• 6:45
  listening to that or
• 6:46
  that's the same thing that i'm looking at or reading so this
• 6:51
  had a resonance that i
• 6:53
  wasn't expecting but now that i think about it obviously
• 6:57
  everybody's thinking where
• 6:58
  SPEAKER_01: do i get information that is meaningful to the work i'm doing
• 7:03
  The other thing that I'm interested in knowing more about is
• 7:07
  that have you noticed
• 7:09
  SPEAKER_00: any change in terms of people's interactions with Gen AI?
• 7:14
  SPEAKER_00: I mean, going beyond just asking questions and expecting answers from AI.
• 7:20
  SPEAKER_01: Definitely.
• 7:21
  SPEAKER_01: We're definitely in the conversation about feedback.
• 7:24
  SPEAKER_01: That was, how do I build a bot that has my voice?
• 7:29
  SPEAKER_01: How do I direct it?
• 7:30
  SPEAKER_01: How do I connect it to multiple sources?
• 7:33
  SPEAKER_01: All of that was very obviously kind of front and center.
• 7:36
  The second piece,
• 7:37
  and this is something that I'm very interested in,
• 7:41
  is that...
• 7:44
  SPEAKER_01: The companies, AI was a ripe area for new companies to enter.
• 7:52
  SPEAKER_01: But now that Google has started figuring things out,
• 7:56
  SPEAKER_01: they are activating AI throughout their school networks.
• 8:04
  SPEAKER_01: And they're putting both Gemini and Notebook LM,
• 8:09
  SPEAKER_01: very pronounced,
• 8:11
  SPEAKER_01: inside all of Google for education.
• 8:15
  SPEAKER_01: websites.
• 8:16
  And so they immediately turned Gemini from something that was
• 8:21
  lagging behind
• 8:22
  SPEAKER_01: ChatGPT and it still is from a user perspective.
• 8:24
  SPEAKER_01: But for the next generation,
• 8:26
  SPEAKER_01: because they all have Chromebooks,
• 8:28
  SPEAKER_01: not all,
• 8:29
  SPEAKER_01: but a huge portion of the United States works on Chromebooks,
• 8:34
  SPEAKER_01: belongs to a Google education platform.
• 8:38
  SPEAKER_01: Now they're getting kids used to Notebook LM and Gemini and
• 8:41
  SPEAKER_01: And my suspicion is that that would set behavior trends into the future.
• 8:48
  SPEAKER_00: They also offer the one-year free use of Gemini for all college students.
• 8:54
  SPEAKER_00: So all of these things are aligned, yeah.
• 8:57
  SPEAKER_01: Yeah, and so it is very clear that now there's a big play.
• 9:01
  SPEAKER_00: Mm-hmm.
• 9:03
  And ChatGPT and OpenAI have a very tall hill to climb because
• 9:09
  they are definitely
• 9:10
  SPEAKER_01: in charge right now and they have tremendous power.
• 9:13
  But will that last with the next generation of young people
• 9:17
  coming in already
• 9:18
  SPEAKER_01: knowing Notebook Alarm,
• 9:22
  SPEAKER_01: Gemini,
• 9:23
  SPEAKER_01: and everything Google?
• 9:25
  SPEAKER_01: It seems like they're set up.
• 9:27
  SPEAKER_01: or at least have a potential to be set for a very successful future.
• 9:31
  SPEAKER_01: And it's not surprising.
• 9:32
  SPEAKER_01: I mean, they're a big company.
• 9:33
  SPEAKER_01: That's what they do.
• 9:35
  SPEAKER_00: Yeah.
• 9:36
  SPEAKER_01: All right.
• 9:37
  SPEAKER_01: And you were looking at an app.
• 9:39
  SPEAKER_00: Yeah.
• 9:40
  SPEAKER_00: The name of this app is Future House AI.
• 9:44
  And it's mostly in the field of,
• 9:47
  I would say,
• 9:47
  biology and medical sciences.
• 9:50
  SPEAKER_00: But it's a very interesting gap.
• 9:54
  The gaps that made the developers of that app to work on that
• 10:00
  is that more papers
• 10:01
  SPEAKER_00: are published every year,
• 10:02
  SPEAKER_00: but fewer breakthrough discoveries are made,
• 10:05
  and scientists have spent countless hours reading literature
• 10:09
  and talking to
• 10:10
  SPEAKER_00: experts,
• 10:11
  SPEAKER_00: and traditional research methods are becoming less efficient.
• 10:15
  SPEAKER_00: So all of these things made them think of developing a new app,
• 10:20
  SPEAKER_00: AI-powered app,
• 10:22
  SPEAKER_00: to do the literature review,
• 10:23
  SPEAKER_00: to help them generate hypotheses,
• 10:26
  SPEAKER_00: and finally,
• 10:28
  SPEAKER_00: actually designing an experiment.
• 10:30
  SPEAKER_01: And for me, this is part of the promise of AI.
• 10:34
  When we started with AI,
• 10:35
  one of the promises was this is going to make humanity better.
• 10:40
  SPEAKER_01: This is going to help.
• 10:41
  And what Future House is coming from is a group of AI
• 10:45
  scientists that decided to go
• 10:47
  SPEAKER_01: the non-profit,
• 10:48
  SPEAKER_01: true non-profit route.
• 10:50
  The open AI were pretending we're non-profit while we're making
• 10:54
  billions of dollars
• 10:56
  SPEAKER_01: and asking for trillions of dollars.
• 10:58
  SPEAKER_01: So...
• 11:00
  They went a totally different route saying we will produce this
• 11:03
  and it will be
• 11:04
  exclusively a non-profit building it for scientists to be able
• 11:09
  to move things
• 11:10
  SPEAKER_01: forward.
• 11:11
  SPEAKER_00: Yeah, and they call their app a kind of AI scientist.
• 11:17
  SPEAKER_00: And as they said,
• 11:18
  SPEAKER_00: their aim is automate parts of the scientific discovery process,
• 11:22
  SPEAKER_00: accelerate breakthroughs,
• 11:23
  SPEAKER_00: and broaden access to science.
• 11:25
  SPEAKER_00: So they actually have multiple agents in their platform.
• 11:30
  And I told you I watched one of their YouTube videos where
• 11:34
  developers talk about
• 11:36
  SPEAKER_00: their platform,
• 11:37
  SPEAKER_00: and their platform is currently under maintenance.
• 11:40
  SPEAKER_00: So it will be available in a week, they said.
• 11:43
  SPEAKER_00: But they used an example of a specific syndrome,
• 11:46
  SPEAKER_00: and they said,
• 11:47
  SPEAKER_00: imagine that you are not an expert in the field.
• 11:50
  You can start with one of the agents,
• 11:54
  and all of the agents are the name of the birds.
• 11:57
  So you can start with Falcon and to do a deep literature search
• 12:02
  and then ask your question.
• 12:05
  SPEAKER_00: And then the system automatically generates search terms.
• 12:10
  and pulls different papers,
• 12:13
  clinical trial data,
• 12:15
  and metadata about resources.
• 12:17
  SPEAKER_00: And then you go to another agent for a targeted question.
• 12:21
  SPEAKER_00: For example, what are the key genes identified in that kind of syndrome?
• 12:26
  SPEAKER_00: And you have another agent to ask another question.
• 12:31
  SPEAKER_00: There is another agent, which its name is OWL.
• 12:35
  And you ask OWL,
• 12:37
  has anybody done a specific procedure for doing something about
• 12:41
  this syndrome?
• 12:42
  SPEAKER_00: And finally, there is another agent, and the name of that is Phoenix.
• 12:47
  And you can ask proposed three novel compounds to treat a
• 12:51
  disease caused by that syndrome.
• 12:53
  So in a minute,
• 12:54
  you go from little domain knowledge to identifying a critical
• 12:59
  gap in the field.
• 13:01
  And what I love about this is that you have an expert at your
• 13:08
  disposal as a series
• 13:09
  of agents that you can have conversations with as a way to
• 13:13
  start the thinking.
• 13:14
  SPEAKER_01: I don't think that's where the thinking stops.
• 13:16
  SPEAKER_02: Mm-hmm.
• 13:17
  SPEAKER_01: And more than that,
• 13:18
  I really don't think you can start as a total novice and with a
• 13:23
  few agents find the
• 13:25
  SPEAKER_01: next cure for a communicable disease.
• 13:28
  SPEAKER_01: I don't think that's where it's at.
• 13:30
  SPEAKER_01: So I think that example, while they give it, is really extreme.
• 13:36
  SPEAKER_01: You need knowledge and you need to understand the mechanisms.
• 13:39
  SPEAKER_00: I agree, yeah.
• 13:40
  SPEAKER_01: And you need to be able to read the literature.
• 13:43
  SPEAKER_00: Yeah, yeah, yeah, I agree.
• 13:44
  Because I was kind of worried when they say you go from no
• 13:48
  knowledge to designing an experiment.
• 13:50
  SPEAKER_00: And we are talking about medical sciences.
• 13:53
  SPEAKER_00: I think it's something serious.
• 13:55
  But the other thing or the other features about their platform
• 13:59
  is that their agents
• 14:00
  SPEAKER_00: use full text papers,
• 14:02
  SPEAKER_00: not just abstract.
• 14:04
  SPEAKER_00: They track citations.
• 14:06
  SPEAKER_00: They know provenance of journals.
• 14:08
  SPEAKER_00: They evaluate the quality of sources.
• 14:11
  SPEAKER_00: And also they provide transparent reasoning.
• 14:14
  SPEAKER_00: I mean,
• 14:14
  you can click into the chain of decisions to understand why a
• 14:19
  paper was selected,
• 14:20
  SPEAKER_00: what evidence was.
• 14:23
  SPEAKER_00: have been used for pulling out pulling up those papers and came into that
• 14:28
  conclusion and they also offer an api for their system so if
• 14:35
  you are working in a
• 14:35
  lab and if you already have a kind of software for doing
• 14:39
  literature review you can
• 14:40
  SPEAKER_00: use their api yeah to do the literature review
• 14:45
  All right,
• 14:46
  so that's another tool in the long list of research tools we've
• 14:51
  been talking about.
• 14:52
  This is obviously,
• 14:54
  as you said,
• 14:54
  aimed at biochemical engineering and medicine and all of that.
• 15:01
  SPEAKER_02: Medicine and biology, yeah.
• 15:03
  SPEAKER_01: When it comes back up,
• 15:05
  I'm really looking forward to trying out what it can do in our
• 15:10
  kind of research.
• 15:11
  SPEAKER_01: Is there really a potential for social kind of research,
• 15:15
  SPEAKER_01: or is it really just biochem,
• 15:17
  SPEAKER_01: which is very important,
• 15:18
  SPEAKER_01: but not my deal.
• 15:19
  SPEAKER_02: Yeah.
• 15:22
  SPEAKER_01: All right.
• 15:23
  So the last thing I want to say,
• 15:25
  just as a teaser,
• 15:27
  we're going to talk about this next time.
• 15:29
  SPEAKER_01: But if you haven't looked at GrammarLilu lately, take a look.
• 15:33
  SPEAKER_01: And this is something that did come up at the conference.
• 15:35
  SPEAKER_01: And that is,
• 15:36
  it was very clear different presenters have been off of their
• 15:39
  system for a few
• 15:40
  SPEAKER_01: days.
• 15:40
  And it already looked different,
• 15:42
  acted different,
• 15:43
  and had two new options in it.
• 15:45
  SPEAKER_01: This is changing constantly,
• 15:47
  and if you're using some apps or you've used the apps and they
• 15:51
  didn't work for you,
• 15:52
  SPEAKER_01: I think about Notebook LM often.
• 15:54
  SPEAKER_01: I loved it from the beginning, but the new features are fantastic.
• 15:58
  SPEAKER_02: Yeah.
• 15:59
  And so take a look at Grammarly if you have a few minutes,
• 16:02
  and we'll see you next time.
• 16:04
  SPEAKER_00: Yep.
• 16:04
  SPEAKER_01: On Azadeh and Guy.
• 16:06
  SPEAKER_00: On AI.