YT YouTube2Text

YouTube2Text → Transcripts → Session 2 AI Based Learning Technologies Fall 2026

Session 2 AI Based Learning Technologies Fall 2026 — Transcript

by Life After AI · 8,903 words · 1,392 segments · language en · Watch on YouTube

Open in transcriber Summarize with AI Download SRT/VTT

Full transcript

0:00Okay, I'm just starting the recording.
0:02So, welcome everybody to the second
0:03session of AI based learning
0:05technologies. I am your facilitator,
0:08your instructor, Joyce Lynn Kelly. Today
0:12we're going to go a little bit further
0:15into the pair framework and look at
0:19reflection
0:21that includes assessment and evaluation.
0:24So we're going to first before we
0:28actually move forward is reflect. Let's
0:31reflect a little bit on last week.
0:35So last week our homework was to explore
0:39prompting using the models that were
0:43provided. So that was just giving you a
0:46little bit more meat to work with in
0:48your prompting technique that could
0:51intersect with your current work. And
0:54the task was to
0:57explore your capstone project. So I do
1:01see that you have responded in the
1:03prompting discussion post. If you have
1:06not yet had a chance to include your
1:09prompting discussion for last week, you
1:11you can still add that. It it is a
1:14requirement three discussion posts, one
1:17for each week and your capstone project
1:20to pass this course.
1:23Any questions or any thoughts you'd like
1:26to add about the prompting exercise that
1:30you contributed to last week? Any tips
1:33or tricks that you'd like to share?
1:38I do see a lot of different um
1:42applications used like perplexity um
1:46co-pilot as well. So that's good to
1:49explore.
1:56Okay, I'm going to share my screen.
2:05We'll look a little bit further into
2:08your prompting. So, because you've
2:10started the process with your prompting
2:12last week and you've started exploring
2:16what your potential capstone project can
2:19be, we will take it a step further and
2:23we will now look at uh evaluation.
2:39Okay. So, if you have not yet had a
2:41chance to log into Canvas and you'd like
2:44to follow our presentation this week,
2:47feel free to log into Canvas module 2,
2:51the gamma presentation that I'm sharing
2:53now. You can follow along as well.
2:57We'll start with a land acknowledgement.
2:59On this National Day for Truth and
3:01Reconciliation, we honor the children
3:03who never returned home from residential
3:05schools, survivors, their families, and
3:08communities. Wherever you are, I
3:11encourage you to learn about the
3:12indigenous lands you're on, and listen
3:14to indigenous voices. As we explore AI
3:18and education, let's remember that
3:20technology reflects human values,
3:23including biases.
3:26May we use this tool to amplif a amplify
3:29under reppresented voices and honor
3:31indigenous knowledge systems.
3:33Reconciliation is ongoing. If you'd like
3:36to act, consider learning whose
3:39traditional territory you're on. And
3:40[clears throat] I' I've included a link
3:42for our home and native land
3:45supporting indigenousled tech
3:47initiatives and exploring how AI can
3:50serve equity. Let's move forward with
3:53both innovation and respect.
4:02Looking a little bit deeper into the
4:04pair framework. So this is your cyclical
4:08framework that I use as a basis as a
4:11foundation for when you're using AI
4:14based learning technologies. And if
4:16you've been using AI based learning
4:18technologies over the past couple of
4:20years, you'll see how many iterations
4:23are improved and different models are
4:25added. And so this model pair is a great
4:31framework for you to evaluate processes
4:35and for you to have a little bit of
4:38confidence and build some trust in some
4:40of these systems by using um a
4:43measurable tangible framework. And so
4:47last week we looked at your problem. We
4:49were formulating the problem to
4:52establish what your what your potential
4:55capstone project could be and then we
4:58explored some tools and in the breakout
5:02session we did look at I did give you
5:04four basic large language models like
5:07copilot uh chat GPTSE for example
5:12and then that AI tools repository that
5:15had a myriad of uh various tools to
5:18explore. for
5:20with the interaction,
5:23we will now focus on the reflection. So
5:28when we're looking at the uh prompting
5:32so when we're looking at our prompting
5:34discussion, we're talking about the I
5:38know many people talked about the first
5:41prompt being insufficient. This first um
5:46the first prompt that I put in did not
5:49give me a good idea or ideas for my
5:52framework for my capstone, but I was
5:54able to talk to the to the AI a little
5:57bit further. So expanding your um
6:01prompting was a reflective action that
6:04helped you to discover that um you'll
6:07get better responses once you start
6:10iterating what you've worked on.
6:14This pair framework had two options
6:17which was the academic option as well as
6:20the marketing campaign option.
6:25Just moving forward this class today. So
6:29our focus for this class will be looking
6:32at assessment reflection and that
6:36includes bias and AI. So I do have a
6:39short video small video I'll share with
6:41you. Make sure the sound is working. And
6:45as you watch this, this one's a couple
6:48of years old. Um, but what I want you to
6:54note is anything that you've experienced
6:59as you've been using generative AI in
7:02these past couple of years and any
7:04biases that you may have encountered in
7:08your um, AI responses.
7:13Give me a second while I make sure the
7:14sound is on.
7:28Are this okay?
7:31[music]
7:39Come with me
7:43and you'll be
7:46in a world of pure imagination.
7:50[singing]
7:52[music] Take a look
7:54and you see
7:58into [music and singing] your
8:00imagination.
8:03I will [music and singing] see
8:06will defy
8:10explanation.
8:11[music]
8:16[music]
8:21[music]
8:23I want to give paradise. [music]
8:27Simply look around and view it. [music]
8:34Anything you want to [music] do,
8:38want to change the world, there's
8:41nothing to it.
8:50There [singing] is [music] no life I
8:53know to compare [music] with your
8:57imagination.
8:59Living there you'll be free.
9:04If you truly [singing]
9:06wish [music] to
9:10be
9:18>> [music]
9:25>> Thank you for watching. Any comments or
9:28questions?
9:29Have you seen that video before?
9:34So, this one's interesting.
9:36I've first started sharing this a couple
9:39of years ago, approaching 2025 when
9:42generative AI had first come out. the
9:46image generation capacity of generative
9:49AI was quite limited and was not very
9:52diversified. So if I was creating images
9:55of a classroom and I'd ask for a
9:59classroom full of students, I'd really
10:01have to reflect and analyze those images
10:05to make sure that the classroom image
10:09was something that was reflective of
10:11real world and was it? Sometimes it
10:13wasn't. there was bias that had appeared
10:16in the image generation. And over the
10:19years, the image generation has become
10:23more improved to include more diverse uh
10:28populations, but there are still issues
10:31with under represented populations
10:35occurring in data sets, and that's in AI
10:38data sets.
10:40So when we're looking at an ethical
10:42dilemma here, there is bias in AI and AI
10:46bias refers to systematic and unfair
10:48discrepancies that are often found in
10:50outputs produced by AI systems. This
10:53bias can manifest in various ways
10:56depending on the data algorithms and
10:59design choices and design choices used
11:01to develop these systems.
11:06I'll show you the actual picture. This
11:08one looks smaller when I click it.
11:11So this is an image of
11:15um inequality and discrimination in the
11:18design and use of AI in healthcare
11:20applications. And this is a cyclical
11:25pattern of discrimination that shows I'm
11:28going to start here. So this is a
11:31realworld pattern of health inequality
11:33and discrimination.
11:35This pattern gets turned into
11:39discriminatory data which influenced
11:44which then influences
11:47design and deployment practices because
11:49they're using discriminatory data and
11:53that leads to application injustices
11:57which then turns into real world
12:00patterns of health inequality and
12:02discrimination. and the cycle will then
12:05continue. This is an example of how bad
12:09data can be um
12:14constantly
12:16um put back into the system. And so we
12:19want to ensure that the human oversight
12:22can filter out some of the bad
12:25discriminatory data to ensure that
12:29under reppresented groups are um still
12:33have a voice and are not lost in biases
12:37that are produced by systems like AI.
12:42minimizing bias will be critical if AI
12:46um if AI is to reach its potential and
12:49increase people's trust in the system.
12:51So there's six different ways to
12:54consider to move forward
12:58in creations of policies of how to
13:01minimize bias and um discriminatory
13:06data. The first is be aware of contexts
13:09which AI can help correct bias for those
13:12in which there is high risk for AI to
13:15exacerbate that bias.
13:18Two is to establish a process and
13:20practice to test and mitigate for the
13:22bias.
13:24Three is engage in factbased
13:26conversations.
13:28Four is f fully explore how humans and
13:31machines can work together.
13:34And I'm noticing here these are all
13:36these are have all human capacities.
13:40Five is invest more in biased research
13:43and make more data available.
13:47And six is invest more in diversity in
13:50diversifying the AI field itself.
13:53So these are issues that happen in
13:57healthcare using discriminatory data.
14:00And additionally keep in mind that when
14:03you use AI systems, many of the systems
14:06are using your data for the training of
14:10their own knowledge bases as well. So
14:12this is why it's important to reflect on
14:16the work that you that you are using
14:19that you are outputting with AI and also
14:24ensure that you are exploring these ways
14:28to help build trust in the system when
14:31you are using AI.
14:38There's an active link here in your
14:41slides. Feel free to look at that in
14:45more detail on your own time. Just as a
14:47summary, these are AI and assessment
14:51strategies that you can start to think
14:54about how to implement or um create
14:58lesson plans or start to shift your own
15:01work in your classroom. These are some
15:04strategies to help you
15:07start to implement AI literacy in your
15:10students as well. So design goals. Here
15:13are some design goals of examples. So
15:17you want to clarify your AI use. What
15:19could you do? You can create a syllabus
15:22segment and an AI use rubric. Why it
15:25matters? It reduces ambiguity and
15:28supports integrity. Sometimes
15:31if you don't include an AI use
15:33statement, students are going to use it
15:35and then you wonder, did they use AI and
15:38it wasn't clear and so students should
15:40know whether they can use AI or not.
15:44Reveal the process. Scaffolded drafts
15:46with educator feedback. This is an
15:49example of um the capstone the capstone
15:54exemplar that I had posted for the
15:57nursing student. She had a revealed
15:59process with scaffolded drafts that
16:02could help teach her nursing students.
16:06Contextualized judgment, case studies,
16:09and local examples requires domain
16:12knowledge and situated thinking. We're
16:13going to explore some of that a little
16:15bit further today as a group um so that
16:18you get a little bit more handson
16:20knowledge and skills with this judgment
16:25and AI literacy and student agency
16:28critique and validate AI outputs. We've
16:30been doing that in the past week. And
16:32then personalized reflection tasks.
16:35These are important when you're
16:37considering how to implement AI into
16:40your classroom and how to do it in a way
16:42that is transparent for your classroom,
16:45transparent for your students,
16:46transparent in your workplace.
16:50And I'll give you some examples as well
16:52of of this. So foundational skills for
16:55educators in this space. I think it's
16:59important to have an AI policy design.
17:02This is clear articulations of where
17:05when AI is allowed.
17:09Prompt engineering for learning. We've
17:11been doing that. Task decomposition. So
17:14break the assessments into meaningful
17:16process, artifacts, drafts, revisions,
17:19reflections. This breaks it down. That
17:21way you can take it step by step how AI
17:25was used. Sometimes you get a a writing
17:28assignment back from a student and the
17:30whole thing just looks like it's AI,
17:33like a copy and paste. If you break it
17:35down a little bit, it gives the student
17:37an opportunity to reflect back on their
17:39work using AI um strategies,
17:46rubric development. I have some rubrics
17:49and we'll we'll actually look at a
17:52couple of AI applications that are free
17:55that can help you also generate rubrics
17:57if that's the space you're looking for
17:59to help define how to use uh AI in your
18:04assessment or in your lesson plans
18:07and then AI output evaluation teaching
18:10learners to identify hallucinations
18:12ver verify sources and asset bias. We're
18:15going we're going to do that today in a
18:18in a small breakout as well.
18:21Okay. Risk limitations and tradeoffs.
18:23I've included the link for how how you
18:26can use um these strategies for you if
18:29you want to look a little bit deeper.
18:36I do want you to take a peek at this.
18:38So, this is a good framework just to
18:40compare policies versus guidelines.
18:48Is anybody using a policy or a guideline
18:51in their classroom currently?
18:57Feel free to unmute your mic if you'd
18:59like to share. If you're not, if this is
19:02new to you, Ann.
19:05Well, we do like I don't teach I work in
19:07a hospital, but we do have an AI
19:11framework for the use of AI in the
19:15hospital. So, it's very like you know
19:18look like there's boundaries and like
19:21access is limited. But I work in uh
19:25development and learning and I just
19:27wrote like I have a an off-site meeting
19:30with my team on Friday and actually I
19:32just wrote down we because we're
19:34reviewing our workflow
19:36to um see how we're changing the roles
19:40of the people in the teams because we're
19:42now using AI more and more and I want to
19:44make sure that we all agree on
19:47what the new workflow looks like and uh
19:51you know when and where we're going
19:53use AI and disclose that to our like
19:56clients in the hospital if you will. So
19:58anyway, so very timely.
20:00>> That's that's good. And I've been
20:02reflecting I' I've been to my doctor's
20:04office recently and on her wall it says
20:07we will be using AI to analyze this to
20:10analyze that. So
20:12>> they've implemented a lot of that as
20:14well trying to be transparent with that
20:16process. Mhm.
20:19>> Thank you, Ann.
20:20>> Okay.
20:21>> And Ambin says, "I don't teach a class
20:24but have developed an AI usage policy
20:26for the organization." That's good.
20:30And
20:31this is an example of guidelines versus
20:36the policies that can help you in
20:39managing AI use. And I know in some
20:41organizations there are fire firewalls
20:44that have already been implemented where
20:48and I should have asked this maybe in
20:50the first class but is there anybody
20:51here who cannot access an LLM because
20:55they're work using their work laptops or
20:57is everyone okay to access chat GPT
21:02um co-pilot and there are enterprise
21:05resources available um through
21:08organizations that have allowed it. So
21:10some organizations will strictly only
21:13grant access to for example copilot
21:16which is under their internet. So it
21:19looks like everybody has access. So
21:21that's good. But there are as I
21:22mentioned policies in action for some
21:25organizations that sometimes they just
21:28block off uh the access. So these are
21:32good uh strategies to support how to use
21:36AI within your organization.
21:41Okay, it's 6:23 right now. I want you to
21:44I'm going to put you into groups and I
21:46want to explore this activity. This is
21:50called AI is wrong. So AI can make
21:54mistakes and responses can be false in
21:57some instances. Maybe that it makes it
22:00more humanlike. What do you do when you
22:03find out it was wrong? Presumably you
22:05have a more reputable resource. What is
22:08it and what makes it more reputable?
22:12Choose an article. So, I know there's
22:14many of you from different parts of the
22:16globe. We have uh Iceland, Turkey, Costa
22:20Rica, um Canada, the West Coast, and the
22:24East Coast. So wherever you are, choose
22:27a reputable news source
22:31and I want you to compare it with your
22:36large language model. So whether you're
22:38choosing Chat, GPT, Copilot, DeepSeek,
22:42choose an article from an reputable news
22:44source and use an AI tool to summarize
22:47it. Compare the AI generated summary
22:50with the with the original article.
22:53Discuss any discrepancies or
22:56inaccuracies in the summary. What might
22:58have caused these inaccuracies?
23:01When you're asking for a summary, you
23:05can either take the link of the article.
23:09Sometimes that works, sometimes it
23:10doesn't, or you'll need to cut and or
23:13copy and paste the entire contents of
23:16the article itself to ask for the
23:18summary. So try either the link in
23:22asking for the summary. If that doesn't
23:24work, try the um try the copy and paste.
23:30So for this discussion, look for an
23:31instance which you think AI has given
23:34you a wrong or bad questionable
23:37information.
23:39Tell it it was wrong and how you know
23:41it. What was the response and what did
23:45you learn? Okay. So, as I mentioned, I
23:48want you to think about
23:51um a reputable news source wherever you
23:53are. I'm going to put you in groups of
23:56about three to four people per room to
23:59explore this activity together. And if
24:02one person could share their screen and
24:05another person uh look up articles to
24:10compare or a summary and then we'll come
24:14back in about
24:1715 minutes to discuss your findings. Is
24:20that enough time do you think?
24:23>> Yeah.
24:23>> 15 minutes is good to compare an
24:25article. Okay. So, we'll do that
24:29and
24:30I will visit each room as well.
25:07same article. Sorry about that.
25:09>> No worries. We put the same article um
25:12on all three platforms and somehow the
25:15language was different. Um the
25:17perspectives were also a little bit
25:19different. I mean this is what I what we
25:21noticed was so for example the one in
25:24co-pilot it started with Trump's
25:26tariffs. It was about tariffs. I mean it
25:28was second one started with the company
25:31uh which was involved and the third one
25:33was I think with Carney right Sigment if
25:35I'm not wrong so the language was
25:38different there were there was a use of
25:40accused and blame and then uh what
25:42Sigman did was he he again said on chat
25:46GPT that summarize so chat GPT
25:50summarized it again but in a different
25:53way so three times you know He tried it
25:58and all three times it was different. Um
26:00the content was probably the same but
26:03the words and the the details were
26:06different. Um and then he asked why are
26:11you using blamed instead of accused? So
26:14we we did some different prompts in it
26:17and the language used is very different.
26:20uh it's I think I think the words that
26:24you use shifts the shifts the
26:26perspective when somebody's looking at
26:29so looking at the output that comes out
26:32of an AI tool.
26:34>> Thank you Embryum. I did include in the
26:37chat the article that that group was
26:40comparing. And so that was an
26:42interesting group to visit because of
26:44the language that the summary of each
26:48lang large language model was choosing.
26:52Um
26:53and so that could alter your voice and
26:55your perspective. So that was a very
26:58good article to compare. Thank you.
27:03Um, I wanted to ask and I don't mean to
27:05put you on the spot, Sigman, but you had
27:06mentioned about Deepseek when you were
27:10using DeepSeek and how you felt about
27:12its responses.
27:15>> Oh, yeah. Absolutely. Deepseek was free
27:19and it started giving me inaccurate
27:22responses
27:24uh and then just refused to admit that
27:27it was wrong. I was asking about the new
27:29software update Golden Gate for Apple
27:33and it said the software update didn't
27:35exist and kept bringing me back to the
27:37bridge in San Francisco.
27:39>> So I argued with it for 10 minutes and
27:41then I just I don't ever use it again.
27:45[laughter]
27:47One thing I want to point out about uh
27:49large language models and it depends
27:51which large language model you're you're
27:53using is it may have a a knowledge
27:56cutoff date but many large language
27:59models as well now integrate the
28:02internet web search. So it can search
28:06recent articles if you have that feature
28:10turned on. And so that's a feature that
28:12you can find in settings and in chat GPT
28:16it's under settings and personalization
28:19and then you'll find a little web
28:20search. If that's not on it may have a
28:23knowledge um
28:26a a knowledge cutoff date which won't
28:29find any recent articles if you're
28:32looking for something that happened
28:33today for example. Thank you for sharing
28:36that segment. Another point that I
28:39wanted to uh show was uh I think that
28:43group had been using claude. They
28:45explored claude and that large language
28:48model did not like links. And so it's
28:51interesting to see how each uh large
28:55language model needs to input your user
28:59in uh your user data.
29:02Thank you. And the next group please
29:09go ahead Anik. Thank you for
29:10volunteering.
29:11>> [clears throat]
29:11>> Um in our group uh we also picked an
29:14article about Trump uh and uh what we
29:19realized we only used it's wise to use
29:21multiple uh LLMs like to compare but we
29:26only used uh chat GPD and um we asked
29:29for the summary and we realized that the
29:31summary was longer than the [laughter]
29:33the article itself. So, uh, you know,
29:37that was surprising. Uh, and then, um,
29:42well, we I don't know, we felt like
29:43maybe there was a tone to the summary
29:46like that didn't like that was more like
29:50um,
29:52um, an opinion than a summary, you know
29:55what I mean? So, but I we're not totally
29:57sure because we didn't like, you know,
29:59had the time to really read carefully
30:01and like do the comparison. Um what else
30:05was uh did we find?
30:10Uh I can't remember. Oh, we think also
30:13like you know one of the reason it's
30:15like the summary was longer than the
30:18article. It's because it wasn't just
30:20summarizing the article, but it was sort
30:23of maybe summarizing like taking
30:26information from all the links embedded
30:28into that article. So it it didn't like
30:31you know like how it provides like um in
30:34the summary it was like first paragraph
30:36talks about this and it was showing like
30:39the link like you know like the the the
30:41source right so there were multiple
30:44sources and not all from that article
30:47you know what I mean so it embedded like
30:49>> so it wasn't a summary basically yes it
30:52analyzed it exactly more than a summary
30:54but maybe it's also because our prompt
30:57uh you know was we didn't spend time on
30:59the prompt you know, so there's that.
31:02Uh, as well, we didn't have time to tell
31:04him, hey, you're wrong. You know, the
31:06summary, like let's say we would have
31:08said the summary is uh is longer than
31:11the article itself. What's up? You
31:14didn't have time to to do that. That
31:16would have been uh fun to do.
31:19>> Thank you for that uh summary. And yes,
31:22it is fun to tell it it's wrong because
31:25I find that it just agrees with you. Oh,
31:27yes, you're right. There are three Rs in
31:30strawberry or however many.
31:33>> H yeah, I I I I I still haven't done
31:36that. I haven't played with AI like
31:38this, you know, but it's good to know
31:40that AI can tell me like can accept that
31:42I tell it it's wrong and you know and
31:44that's I'm joking here like just so you
31:46know I can agree. I love AI already. Oh
31:50that's great. I'm kidding. I'm kidding.
31:52I'm joking. [laughter]
31:54Another point that Anek had made is the
31:57sources that had came up in your summary
32:00when you asked it to summarize and it
32:02came up with a few other articles that
32:04you didn't include. Check those sources.
32:07Sometimes
32:08>> they don't have the information that was
32:11summarized or it's a dead source. It
32:14doesn't even exist. The link doesn't
32:17activate. And so some of some of those
32:20sources are hallucinations in it in it
32:22of themselves.
32:24>> Huh. Interesting. Okay. Could it also be
32:27like that it's
32:29>> I don't know that the developer of the
32:31tool like let's say anthropic is also
32:34trying like you know they have their
32:36like things that they want to push to
32:39us.
32:41I that's a theory that I want to
32:43research actually an what I was thinking
32:45about last week was um I'm not sure if
32:48you're looking at Google and the way
32:50that Google works now when you put your
32:52search term in
32:54>> and then it has its AI summary
32:56>> and its AI summary is pulling from
32:58information that it's already
33:00>> reading from websites which keeps you
33:03from actually clicking onto websites. So
33:07the validated source people are skipping
33:10now because it's appearing in an AI
33:12summary. But this practice that we're
33:14doing now gives you the ability to check
33:17that is it still accurate. Do those
33:19sources still exist? And like you said,
33:22is the link or the company is wanting
33:24you to go to that link or
33:27now it's now it's gone. Now the link
33:30ability is summarized. And so people are
33:33accepting these summaries as as what
33:36they see.
33:38>> I don't want to know. I don't want to
33:39hear like I preferred before like you
33:42know discourse when I didn't ask myself
33:45about AI and I was living in La La Land
33:48everything.
33:49>> It's a good question. Yes. I see it all
33:51the time like is that AI?
33:54>> Thank you. And we have a question from
33:56Andrea.
33:58Hello Andrea.
33:59>> I think hi. No, I think uh I was going
34:01to tell you that I really like the
34:03activity to do with the students because
34:05we tell them you know you you you have
34:07to be like detectives like uh be be
34:11ready to uh find what is what is not
34:15accurate
34:17because might be that AI doesn't have
34:19the right answer. So instead of copy
34:22paste,
34:22>> yeah, investigate
34:25>> text I uh generative AI is a text
34:29predictor. It's not a fact machine. And
34:32so it's important for students to know
34:35that these are probabilities that the AI
34:37has come out with in your output and
34:39it's not factual. And it even states it
34:43at the bottom of chat GPT. AI can
34:45hallucinate. We can AI can make
34:48mistakes. And still so many people are
34:51putting in an input and saying, "Hey, I
34:53read it there on the Google the Google
34:55AI summary."
34:57Thank you for that uh perspective. Was
35:00there another group that didn't have a
35:01chance to share yet?
35:05Yeah, it was my group. We uh read an
35:08article about
35:10uh indigenousled developments in Canada
35:13as far as uh housing shortage.
35:16uh we didn't have much time to really go
35:18through the summary. We did uh we used
35:20chat uh gpt first of all and it was very
35:24lengthy in length. So that was what an
35:28was mentioning. So we also noticed that
35:30so it took us quite a while just to read
35:32the summary. Um I did input the same URL
35:36into copilot and I've been using copilot
35:40uh recently just at my job for the first
35:41time. I was mostly using chat GPT
35:44before, but I find Copilot is actually a
35:47little bit easier to use as far as their
35:48summaries. There's just a short bulleted
35:50list of points and a key takeaway. It
35:54was like a third of the length of the
35:55one that was generated by the other one.
35:57So, thought that was kind of
35:58interesting. Uh, both were very positive
36:01in their outlook as far as sentiment. I
36:04guess our news article was kind of a
36:06positive one. So,
36:08uh, yeah, I wish we had a little more
36:10time to do the detective work, but those
36:12are some of the things that we noticed.
36:14Yeah,
36:17>> thank you everyone for sharing your
36:19experiences. And just to take it a step
36:22further on your free time, I'd encourage
36:25you to go ahead and try it. Try and tell
36:28AI that it's wrong about [gasps] your
36:31article and see how it responds to you.
36:35So use this as a continuous practice to
36:39evaluate the responses and even the
36:42summaries to make sure because I know at
36:44work you may have um like a meetings
36:47notetaker for example that joins and it
36:50may just automatically summarize your
36:52meetings your meeting minutes but it is
36:55it entirely accurate? Did it capture
36:57what it needed to capture? So having the
37:00human oversight just like in that model
37:03that I showed you with the steps that
37:06those all include human oversight at the
37:09end of the day. So remember not to
37:11outsource your thinking when even when
37:14it comes to summary summaries and
37:16articles like this.
37:19Okay. So what time is it now? We're at
37:226:55. Let me just get back to
37:26the presentation. We're going to take 10
37:28minutes.
37:30If you do have a question, I will stay
37:33behind. Otherwise, feel free to grab a
37:37tea and I will um share the the timer
37:42here.
37:47Okay. So, we'll see you back in 10
37:49minutes.
37:51If you have any questions, I can stay
37:53back.
38:12Hello.
38:14Hello everyone. Welcome back. Got my mic
38:17on.
38:20Okay, let's explore some AI tools and
38:24resources. Outside of large language
38:29models, there are some specialized
38:31tools. And these I don't um I don't have
38:34affiliation with any of these. These are
38:36ones that I've used in the past that
38:39I've um completed an evaluation for. And
38:43so there's tools that you'll come across
38:45that you'll like, tools that you um
38:49would want to keep in your back pocket.
38:51So bookmark the tools that you find
38:53useful.
38:56This one is magic school. So this one is
38:59in video. Each of these you can click.
39:02I'm going to log into
39:05Magic School first. I know a couple of
39:08you have had experience with Magic
39:11School. Is Magic School new for some of
39:14you?
39:16Okay. If you're an educator,
39:19you can access Magic School as a premium
39:23and you can use your Google account, uh,
39:27Microsoft account,
39:29and you'll have access to an AI
39:33assistant, Raina, as well as
39:36professional tools for teaching, tools
39:39for your student, and workflows.
39:42This has a number of different features
39:45here.
39:47I'm going to and here's a custom chatbot
39:49feature that I mentioned to you last
39:51week that you can create your own custom
39:53chatbot using uh Magic School. I want to
39:57direct you to the rubric generator. So,
40:00this can help with assessments. And this
40:03is a type of tool that is uh simple to
40:09create using this AI tool. So, you can
40:12define your grade level. So it goes all
40:14the way from kindergarten. It also has
40:18uh postsecary, secondary school,
40:21professional staff as well.
40:25And then you can
40:28enter your point scale. So how many
40:31point scales do you have on your rubric?
40:33Met, not met, or needs improvement, for
40:37example. You can select that. Here
40:41you can add your files. So, if you have
40:44an existing lesson plan that you want to
40:48create AI literacy and include a
40:53transparent rubric, this is a good tool
40:55to help you create a rubric that your
40:59students can use transparently.
41:03Add your description. So, this one has a
41:05description already. So, write a
41:06persuasive essay that convinces the
41:08reader to change a school policy of your
41:10choosing.
41:12You can also customize it by adding
41:14files here and then generate your
41:17rubric.
41:20Okay, so that's a good example. And
41:22here's an exampler.
41:27Scroll to view the output. So this is an
41:31exemplar of a fivepoint scale for a
41:33ninth grade using SWBAT
41:36writing an argumentative essay.
41:39Um and then the exemplar used this
41:45scale. So excellent, good, satisfactory
41:49and needs improvement. And the thesis
41:52statement criteria presents a clear,
41:54focused and compelling thesis.
41:56And that's excellent. And then needs
41:59improvement has presents a vague or
42:01unclear thesis statement that does not
42:03effectively address the topic or purpose
42:06of the essay. So here's an example of a
42:10rubric that was generated using magic
42:13school.
42:16Okay. So this is a good tool.
42:21I'll go back as well.
42:24There are writing feedback as well as if
42:27you're using YouTube and you want to
42:31generate questions. This is a good one
42:34to
42:35enter the transcript. So you can develop
42:39YouTube questions from
42:41a YouTube video for your students as
42:44well and then select the grade level.
42:52And then I'll go back to the custom
42:55chatbot just so you can see similar to
42:59PO that I showed you last week. The
43:03layout is this is very similar. So
43:05create a chatbot. You give it
43:07instructions.
43:09Do you see this instruction? It's
43:11similar to the framework that I gave you
43:14last week on prompting. So give it a
43:16role, the goal, the results, and the
43:18rules. So this is a good example of
43:21creating custom chat bots using a custom
43:24framework uh for your prompting. And
43:27then I'll just look at the examp ex
43:30exemplar
43:32chat bots in comparison to agents are
43:40if if you're asking about agents and I
43:42know some of you had asked before but
43:44chat bots you engage with. So you ask it
43:46a question and it will give you a
43:48response. Agents is a new AI technology
43:53that will act autonomously on your
43:56behalf and make decisions. So that's the
44:00difference between a custom chatbot and
44:02an agent. There is an agent an agent
44:05functionality in this workflow and it
44:07looks like there's one for assessment.
44:10So, class writing feedback, you can
44:12create an assignment,
44:14enter your details, upload it,
44:19review the rubric that we looked at, and
44:22then send feedback
44:24on the students assignment based on your
44:27rubric. But this, as I mentioned, will
44:31require so much human testing in the
44:34end.
44:35The chatbot is a great first step into
44:38creating a workflow. The chatbot will
44:41let you analyze the response first to
44:44help you gain trust. Okay? And this is
44:47just one tool that we're going to and
44:50this is a tool that I want to you to
44:53analyze even further using an evaluation
44:57framework.
44:59exit out of this.
45:04Um, Ed aid is similar.
45:08This one has more of
45:12an American background. It uses grade
45:15levels in comparison, but it's similar
45:18to
45:19um
45:21the format of Magic School and it is
45:24also a premium application.
45:28And you have tools here for lesson uh
45:31lesson planning.
45:35Uh Socratic seminar. I really like I
45:38know some of your discussion posts have
45:42um you've suggested that your chatbot or
45:46your capstone project could pose
45:48questions to the student and without
45:51giving the responses and have them speak
45:54back to the chatbot. And this is a
45:56Socratic manner and this could be the
45:59tool that you use for a Socratic
46:01seminar.
46:05Okay. So explore some of those features
46:09on edu aade as well as
46:12um magic school
46:16and I'll show you gamma.
46:19Gamma I really like and that's the tool
46:22that I actually use for presentations
46:26and I want you to note that I do not pay
46:29for any of these applications that I'm
46:32sharing with you except for chat GPT. So
46:35all of the applications like Gamma
46:39Ed and Magic School I've used in helping
46:43to create lesson plans as well as
46:47rubrics.
46:49simple chat bots as well. And I I did
46:51not pay for these tools in using them in
46:54my classrooms.
46:55So I encourage you to explore those
46:59tools that you have um access to and see
47:04what you can develop using the free
47:06accounts. See what you can do and keep
47:09those in your back pocket. Keep that as
47:12um on your favorites list. And I did
47:14include the AI educator tool repository
47:18for you here
47:24and a few other links. So this link as
47:28well is for
47:32chat B uh prompts. So it will give you
47:35rubric prompts if you want to create
47:37your own in your own large language
47:39model. So, you're an expert teacher,
47:42curriculum writer, skilled in creating
47:44assessments, and evaluating student
47:46work. Your task is to create and then
47:50enter your subject, your grade,
47:53and there's many example prompts that
47:56can help you with creating in different
48:00large language models.
48:02So, that's a really good tool as well.
48:06Okay. other links here that can help in
48:09your assessment
48:11and I want you to view those on your own
48:14time. We won't have so much time to get
48:17through it all today,
48:23but I do want you to take a look for
48:25this
48:28pop out.
48:31So this is an example of a framework.
48:36Let me scroll to the top. This is an
48:39example of a framework for evaluating AI
48:42tools and the tools that I've shared
48:45with you. I have used this framework on
48:49the tools that I like. And so this is
48:53just one example.
48:56Is your link working? Is everybody's
48:57link working for that page? you have
48:59access to this.
49:01Okay.
49:03So, this is another rubric that I like
49:05that helps evaluate AI tools. So,
49:08whether you're looking at gamma for
49:10example, you can take a look at the
49:13terms of use, privacy settings to help
49:17answer these questions. So, some
49:20important factors here is functionality.
49:24Um, and I know we've mentioned this
49:26before, the AI interface is intuitive
49:29with little to no learning curve for new
49:31year for new users. Some users find some
49:36um,
49:38who was it? Was it Tyler? You were
49:39mentioning you like co-pilot better the
49:41way it outputs with the um, the bullet
49:44points versus
49:46>> GPT was a bit lengthy.
49:48>> Copilot. Yeah, I've just found it a
49:50little bit easier to use.
49:52Mhm. [clears throat] And that's an ease
49:54of use.
49:56>> Yeah,
49:56>> that's um that's an important feature
50:00when you're evaluating AI use in your
50:03class classroom. Another one is ethics
50:06and bias, which we're continually
50:09reflecting upon. I'm going to scroll
50:12down here to accessibility.
50:15So, in a classroom like this where we're
50:17online and we have students that are
50:20across the globe, I don't know what your
50:23access is in terms of your demographic
50:25if you have um access or if your
50:29workplace blocks you from accessing
50:31certain language large language models.
50:34That's something you need to establish.
50:36Who is your audience? Who are you
50:38working with to ensure that you have the
50:42right tools and that it's accessible to
50:45your students?
50:47So, make sure that in your classrooms
50:49that the students can not only they not
50:52only like how the user interface is, but
50:55can they access it or do they have to
50:57pay? Those might be restrictions for
51:00their use of using those tools.
51:03and inclusivity. That's another um it's
51:06another consideration. I had written in
51:10one of my blog posts about the internet
51:13access that is restricted in many areas
51:15across Ontario. So, it's important to
51:19know again where are your students? Can
51:21they access is it inclusive for them to
51:23access the tools and the apps that
51:26you're suggesting?
51:29Um, I'm going to scroll down to privacy
51:33and data. So, here, data, privacy, and
51:37ownership. Some of these,
51:41um,
51:42some of these
51:44you won't know until you look into it.
51:47I'm just going to copy and paste
51:50into the chat.
51:56Open AI's policies and their privacy
51:59policies.
52:01In many cases, you may not be even
52:04looking at this. You may not be even
52:06looking to see where your information is
52:09stored, where is um your personal data
52:13used. And so looking at the privacy
52:18policies and understanding any
52:20disclosures,
52:22um any issues. I know that some
52:25some institutions, for example, will
52:28choose not to use Google because it's
52:31American-based and I'm in Canada and so
52:34the um the control and the privacy
52:40goes back to a Canadian tool. So what
52:43are the terms of use? These are
52:46important factors and this
52:50type of framework for evaluation can
52:54help you in defining and presenting to
52:58an organization a policy that can be
53:01utilized with more trust and confidence
53:03in your workplace or in your classrooms.
53:07This is just one example. I'm going to
53:10exit out of this one.
53:16This is one example of an AI tool
53:20that I use for evaluation.
53:24There's another AI tool that I want to
53:26share. And this one is shared as a
53:30private file here. So, if you go to the
53:33next tab, if everyone can follow because
53:35I want you to complete this
53:40um this rubric. So, when you click on
53:44that tab, you'll see copy document and
53:48make a copy.
53:50Can everyone do that?
53:53So, I make a copy and then I have a
53:55brand new page here. I'm just going to
53:58log my email. I know sometimes you might
54:01need permission
54:07to access that. Is everyone able to
54:11access this link?
54:15Is anyone having trouble accessing this
54:17link? Excellent.
54:20Okay, so this is a tool that Let me make
54:24this bigger.
54:32This is a tool that you can use for
54:35evaluation and you can toggle
54:37on and off these
54:42selections.
54:49So for example,
54:51if you're evaluating
54:54um let's say Gamma, Gamma is the
54:57presentation application that I like to
54:59use. The tool fosters a sense of
55:01belonging. I'm going to say I strongly
55:04agree.
55:05And let's say I wanted to change my
55:08mind. I can just click off of this and
55:11it will uncclick it.
55:15Okay. So once I click all these, it will
55:19give me a category grade
55:22whether it's useful or not.
55:28Okay. So what I want you to do and I'll
55:30put you into groups into breakout groups
55:33is explore this framework. So
55:39uh Darlene, you're not getting the same
55:40page.
55:44Did you click the link that was in the
55:47presentation? The clickout?
55:53>> No, I did the one that was in the chat.
55:57Where am I supposed?
55:59>> I just put a link in the chat. Try that.
56:03>> Sorry,
56:04mine doesn't look the same at all. Let
56:08me try again.
56:09>> Okay, so I've just entered the link in
56:12the chat for you there.
56:14So, I'll open up uh breakout breakout
56:17groups. Again, I want you in your
56:19breakout groups to decide on one AI tool
56:23to evaluate and go as far as you can to
56:26evaluate these segments based on what
56:30you can find. So, you may need to look
56:33into the tools terms of service or their
56:36privacy policy.
56:38And then, like I said, click on what you
56:42agree on here. here. And if you change
56:43your mind, just double click it, toggle
56:45it off.
56:47Okay. Are there any questions
56:50about that? I'm just going to give
56:53permission here.
56:59>> Okay.
57:00>> So, I'll open up the breakout rooms. I
57:02do see a few requests for access, so
57:04I'll permit access while I'm opening the
57:08breakout rooms.
57:13Um, we'll look at 20 minutes. 20 minutes
57:17and then we'll come back.
57:23I'll do 25 minutes.
58:02Hello everyone. Welcome back.
58:05I hope you all had a good chance to
58:08explore some of those
58:11um frameworks for evaluation.
58:14Can I ask for a member of each of the
58:17groups to share an update with their
58:20experience and the tool that you chose
58:22to
58:24uh that the tool that you chose to
58:26evaluate? We have three groups
58:31and I'm going to ask for someone aside
58:33from Embream.
58:36[laughter]
58:41>> I can share from that group.
58:43[clears throat]
58:44>> Thank you. T
58:45>> um we didn't get through all of it. Um,
58:48but I think um this may be a little bit
58:51of a tangent, but something that um some
58:54questions I guess that have brought up
58:55for me um is who the the who's the
58:59person who's going to be using the
59:01technology because some of a lot of the
59:03questions that we asked um the answer is
59:06going to be really different depending
59:07on whether you're looking at it from the
59:09student perspective or you're looking at
59:11it from the teacher perspective. Um, so
59:14we looked at at EdUade. Um, and it it
59:18looks like we didn't get all the way
59:20through the checklist, but it looks like
59:22it um requires a lot of teacher
59:25oversight, which I think is a good
59:26thing. Um, at that at you know, at the
59:29end of the day, the the teacher is able
59:31to modify the output um and is
59:34responsible for making the pedagogical
59:36decisions. um we didn't get as far as as
59:40safety and privacy, but again that's
59:42going to look really different depending
59:43on um whether you're the teacher or the
59:46student and the age of the student. So
59:48I've worked primarily in elementary
59:50school, some junior high um sharing of
59:53information is going to look drastically
59:54different for a grade seven than it is
59:56for a university student. Um, so all of
59:59those things are are things that you
1:00:01need to consider. And my experience has
1:00:03been um working for school districts is
1:00:05that our the technology department does
1:00:08this. So they they look at checklists
1:00:11and evaluate all of these things. But I
1:00:13think this is a really valuable tool for
1:00:15for teachers to look at to start
1:00:18thinking about some of these things
1:00:19before they submit the tool to that
1:00:21department for evaluation because maybe,
1:00:25you know, from that preliminary look,
1:00:26we're going to determine that it's it's
1:00:28not a tool that's probably going to be
1:00:29approved anyway. So that let's explore
1:00:31something else before we look for that
1:00:33approval.
1:00:37>> Thank you, Tanya. Does anyone want to
1:00:39add about what Tanya was mentioning? I
1:00:42know um another group had mentioned that
1:00:44they have a similar process with their
1:00:47school that they have an IT department
1:00:49that will typically go through this
1:00:51process. But as Tanya said, it is
1:00:53valuable to have that agency for
1:00:55yourself to evaluate that tool on your
1:00:57own.
1:00:59Thank you. And the next group, please.
1:01:06>> I guess I can speak again for my group.
1:01:09So we Yeah. So we went into uh Grockbot
1:01:13um which is a very cool tool that's um
1:01:17made by XAI. So if you haven't explored
1:01:20that, it's really easy to use. I've put
1:01:22the link in the chat here. So basically
1:01:24you go into Grockbot and within
1:01:27seriously 10 minutes you can have whole
1:01:30army of Grock bots that you create
1:01:33working on your behalf. So, I went into
1:01:35this last night just because I was
1:01:36curious about it and I within like I
1:01:40said 10 15 minutes I had four bots
1:01:42working for me. One was cleaning up my
1:01:44Gmail. The other one was looking for a
1:01:46house in London, Ontario. Another one
1:01:48was looking for instructional design
1:01:50work in the province of Ontario.
1:01:52So, really cool. Uh, our group did talk
1:01:54about the safety. You know, you're
1:01:56providing the bot access to your, you
1:01:59know, personal platform such as Gmail.
1:02:02Um, and how that happens is each bot has
1:02:05their own basically computer screen
1:02:07which you then open and then you can
1:02:09sign into that application on behalf of
1:02:11the bot. So once you sign them into
1:02:13Gmail, you give them prompts and then
1:02:15they can follow your prompts to then uh
1:02:18accomplish different tasks. But yeah,
1:02:20definitely the safety because you're
1:02:21sharing your your login credentials and
1:02:24then what are they doing once they're in
1:02:26your Gmail, right? So, they'd be able to
1:02:29get a lot of data on you as far as what
1:02:31your shopping preferences are on Amazon
1:02:33or a whole bunch of other
1:02:39>> Perfect. Thank you for sharing for
1:02:41sharing that. And that's a I put the
1:02:43link in there as well. And it's a
1:02:47premium access. So there is I think
1:02:51Tyler you mentioned 7 days you can trial
1:02:53this feature and then it's minimum $20
1:02:57afterwards. So when you're triing
1:03:00accounts look for the premium because
1:03:03you may meet uh maximum whether it's
1:03:06tokens or the day limit as well.
1:03:10Thank you for sharing that. And
1:03:12additionally at in Tyler's group, we
1:03:15looked at the privacy that it's by uh X
1:03:20uh SpaceX and that information is also
1:03:23housed in the US.
1:03:27>> Okay.
1:03:28>> I have a I have a comment maybe about
1:03:31the rubric that I'd like to share.
1:03:33>> Yes, thank you Lena. when I was reading
1:03:35the criteria about safety because I
1:03:37guess that somehow to me it is my
1:03:40biggest concern anytime I'm signing in
1:03:43any platform right but um it's the
1:03:46criteria about robust because I find it
1:03:49difficult to know I feel like I'm not
1:03:53equipped to identify if for because it
1:03:57it gives an example here the tool is
1:03:59designed so that I can withstand cyber
1:04:01security attacks
1:04:04But I'm not equipped to do this. So I
1:04:07mean I'm a I I I I like the I mean I I I
1:04:13understand how important to know this
1:04:18uh but I just feel like also how can I
1:04:23develop this skill or this competence
1:04:27>> and some of those privacy issues as well
1:04:31may be not giving away important data
1:04:34information.
1:04:36So holding that like any personal info,
1:04:40credit cards, addresses, names,
1:04:43sometimes you can make that a more
1:04:45highle input rather than including
1:04:49personal details. If that were to be
1:04:51compromised, could an AI or um whatever
1:04:56hacker find out more info if you left it
1:04:58really high level?
1:05:01And so sometimes just omitting personal
1:05:04info is the best way.
1:05:07Thank you. Thank you for sharing those
1:05:09concerns. It can be concerning. And so
1:05:11when I do look at and I learn something
1:05:14every time that we explore this this
1:05:18exercise and it's it's shocking like I
1:05:21think uh Darlene's group we were looking
1:05:24at chat GPT.
1:05:28I think before it cut off, it was saying
1:05:30how it will look at your friends list
1:05:34and send them information if they
1:05:37haven't yet subscribed. So now it's
1:05:39saying once you sign up, we're going to
1:05:41look at your address book. And maybe you
1:05:44didn't even know that, but after reading
1:05:46that, now we know why they get
1:05:48suggestions because they were part of
1:05:50your address book or why you get
1:05:52suggestions for using certain tools that
1:05:55might be part of the privacy policy of
1:05:57these of these applications.
1:06:00Thank you. Is there any other
1:06:02information that you would like to add
1:06:05or any questions or comments about that?
1:06:11Thank you. Thank you for participating.
1:06:13Again, I do encourage you to continually
1:06:17assess your AI tools as well as the
1:06:20outputs. And this I hope you keep in
1:06:24practice
1:06:26continually when when working with your
1:06:28students. Okay. So, I just want to share
1:06:32with you the assessment analysis for
1:06:34next week's discussion. You'll pick in
1:06:38Canvas, you'll pick two different
1:06:43two different questions out of these
1:06:46eight to respond to.
1:06:49And you can respond in Canvas by next
1:06:52Tuesday. If you haven't had a chance yet
1:06:56to complete your prompting discussion
1:06:58from last week, please make sure that
1:07:00you've done that.
1:07:06For next week,
1:07:08this is your presentation week. So for
1:07:11next week, I want you to put into
1:07:13practice the pair the pair framework
1:07:17conceptualizing your capstone project.
1:07:20So for this presentation, you'll share
1:07:23the problem that you have, the AI tool
1:07:27that you chose to use, your um
1:07:31implementation,
1:07:33and your reflection. So the P A IR so
1:07:36far, you can use any tool that you like
1:07:39to share this presentation. As I use
1:07:42Gamma here, I encourage you to try it.
1:07:44If you haven't yet tried Gamma, use the
1:07:46free version. You can use PowerPoint
1:07:50um whatever tool you want to try or even
1:07:54in video try making a brief video.
1:07:57However you find is the best way to
1:08:00present your work. I will not tell you
1:08:03how to present your work. You know the
1:08:05best way for yourself. So I'll give you
1:08:07that uh agency to choose how to present.
1:08:11So like I said, be prepared to present
1:08:13your problem, the AI tool you used, your
1:08:17interaction with that tool, and
1:08:19reflection. It can be done by Wednesday.
1:08:23You can submit it by Wednesday, but you
1:08:25can also use that as a an opportunity to
1:08:29gain feedback from your peers. So if
1:08:32you're stuck on the idea, you can have a
1:08:36few more days. So, Wednesday you'll
1:08:38present and by Sunday you'll uh submit
1:08:41your capstone.
1:08:44Okay. Any questions about your capstone
1:08:46presentation next week?
1:08:50Okay. I'm really looking forward to
1:08:52seeing the problem and the AI tool that
1:08:55you've decided to use. And a reminder
1:08:58again to complete your discussion post.
1:09:03>> I have a question.
1:09:05>> Yes. Hello.
1:09:07>> Just the presentation roughly how long?
1:09:11>> 5 to seven minutes and then you'll have
1:09:14questions after maybe five minutes. So
1:09:16each person will have approximately 10
1:09:18minutes maximum.
1:09:20>> Okay, perfect.
1:09:21>> 10 to 12 minutes maximum.
1:09:23>> Okay, thank you.
1:09:25>> Okay, thank you everyone for your time
1:09:27this week. If you have any questions,
1:09:29feel free to send me a message. I'll
1:09:31stick around for a few moments afterward
1:09:34and I look forward to your presentation
1:09:36next week.
1:09:37>> Thank you.
1:09:38>> Thank you so much. Class is dismissed.
1:09:41>> We'll see you.
1:09:43>> Good night everyone.
1:09:45>> Bye. Good night.
1:09:46>> Good night everyone.
1:09:47>> Good night.
1:09:48>> Good night.

About this transcript

This page contains the full transcript of Session 2 AI Based Learning Technologies Fall 2026 by Life After AI, generated from the public captions YouTube serves with the video. The transcript has 8,903 words across 1,392 segments, with the original timestamps preserved so you can click any line to jump to that moment in the embedded player.

What you can do with it

Use the transcript to take notes, quote the speaker, build a study guide, generate a summary with ChatGPT or Claude via the YouTube Summary tool, or export it as a timed subtitle file with YouTube to SRT. You can also re-open it in the transcriber to choose a caption track available on the video.

Free YouTube transcript tool

YouTube2Text is a free YouTube transcript generator — no signup, no daily limit. Paste any YouTube link and get the full transcript instantly, with timestamps, click-to-jump, caption language selection, AI prompts for ChatGPT, Claude, and Gemini, and exports to TXT, SRT, VTT, or Markdown.