1
00:00:00,000 --> 00:00:02,800
Episode 21. AI as a Tudor.

2
00:00:07,040 --> 00:00:11,040
For generations, students returning home from school were met with a question,

3
00:00:11,040 --> 00:00:15,840
what did you learn in school today? For just as long, the answer has been nothing.

4
00:00:16,480 --> 00:00:21,440
What was lost on all those generations is just how useful that question could be

5
00:00:21,440 --> 00:00:27,200
for our learning. We know that reconstructing what we have learned helps to make it stick.

6
00:00:27,200 --> 00:00:32,320
We know as well that explaining things we know to other people helps us to learn it.

7
00:00:32,320 --> 00:00:37,840
Even if our explanations are incorrect, articulating our understanding can be a valuable

8
00:00:37,840 --> 00:00:42,560
learning activity. If more of us had answered the question, what did you learn in school today,

9
00:00:42,560 --> 00:00:47,360
then our teachers may not have been met with silence when they asked if we had any questions,

10
00:00:47,360 --> 00:00:52,640
and we might remember some advanced algebra today. Who knew that those nosy adults who

11
00:00:52,640 --> 00:00:57,760
were annoying their children were such brilliant learning scientists. Now, it may be too late for

12
00:00:57,760 --> 00:01:03,040
those of us who are adults and some with long gray hair to benefit in school from answering the

13
00:01:03,040 --> 00:01:08,880
question, what did you learn in school today? But we can still use it as a valuable learning tool,

14
00:01:08,880 --> 00:01:15,440
and we can also help our students find a better use of AI than the unsatisfactory, at least to me,

15
00:01:15,440 --> 00:01:20,720
having to do your homework. Some AI tools have been developed specifically for the purpose of

16
00:01:20,720 --> 00:01:27,040
being a tutor. But even those that were not designed for tutoring can be useful. I've used

17
00:01:27,040 --> 00:01:32,640
Generative AI tools trained on large language models to provide a chat companion to help me

18
00:01:32,640 --> 00:01:38,400
explore topics of interest and the potential for this being an agent to whom our students can tell

19
00:01:38,400 --> 00:01:44,080
what they learned today is clear to me. If you open your favorite Generative AI large language

20
00:01:44,080 --> 00:01:49,200
model and add a prompt like, I learned about game theory today, would you like me to tell you what I

21
00:01:49,200 --> 00:01:55,200
learned? Of course, you can substitute any subject that you want. You'll be met with an enthusiastic

22
00:01:55,200 --> 00:02:01,200
partner who prompts you to explain what you know and to correct your errors. This seems a strategy

23
00:02:01,200 --> 00:02:06,480
we can teach our students. Even if the explanations they get from AI aren't perfectly aligned with

24
00:02:06,480 --> 00:02:11,920
the teacher's language and approaches, it is going to explain the material in a different way,

25
00:02:11,920 --> 00:02:17,440
which we know also helps learners. You know, AI is here. It is going to be a game changer in

26
00:02:17,440 --> 00:02:23,280
education, especially if we can find ways to help our students become smarter rather than just helping

27
00:02:23,280 --> 00:02:49,280
our students complete assignments.