1
00:00:00,200 --> 00:00:02,880
And now an exercise for you.

2
00:00:02,920 --> 00:00:07,040
You've just hopefully gone in and run this for other web pages, and you've maybe tried tweaking it

3
00:00:07,040 --> 00:00:08,960
for a different business purpose.

4
00:00:09,480 --> 00:00:18,280
The exercise is now to come in here and see if you can make your own message to GPT from scratch.

5
00:00:18,640 --> 00:00:23,400
So first of all, write a simple system prompt.

6
00:00:23,560 --> 00:00:29,840
Now you should think about something completely different to do here, perhaps still related to to summarizing

7
00:00:29,840 --> 00:00:30,640
information.

8
00:00:30,720 --> 00:00:34,880
So as an example, you might want to summarize the contents of an email.

9
00:00:35,160 --> 00:00:40,760
Or perhaps you might want to take an email and come up with a nice subject, a possible subject for

10
00:00:40,800 --> 00:00:43,960
that email that might be an example of a business task.

11
00:00:44,200 --> 00:00:51,120
So in the system prompt here, describe the task that you want it to do in the user prompt.

12
00:00:51,120 --> 00:00:58,240
You'll want to put in here the contents of the email or the thing to be translated or whatever the is

13
00:00:58,240 --> 00:01:00,120
going to be the prompt with the instruction.

14
00:01:00,880 --> 00:01:04,280
Then you now have to do some typing, just like I did some typing.

15
00:01:04,320 --> 00:01:05,720
It's your turn to do some typing.

16
00:01:05,960 --> 00:01:12,030
Replace this messages equals you're going to want to put in here something which will describe your

17
00:01:12,030 --> 00:01:13,470
messages to OpenAI.

18
00:01:13,630 --> 00:01:18,990
So this is going to be a place where you're going to put in a list of dictionaries, two dictionaries.

19
00:01:18,990 --> 00:01:24,310
And it's going to have some things in it like role and content, and cursor is determined to fill it

20
00:01:24,310 --> 00:01:24,830
in for me.

21
00:01:25,390 --> 00:01:26,230
But don't cheat.

22
00:01:26,430 --> 00:01:29,270
Uh, try and see if you can't do it yourself and then cheat if you can't.

23
00:01:29,310 --> 00:01:31,230
Of course, it's always good to use AI tools.

24
00:01:31,230 --> 00:01:32,510
It's not cheating in the least.

25
00:01:32,910 --> 00:01:41,710
And then you have to uncomment this line and have response equals and type in the the the command that

26
00:01:41,710 --> 00:01:46,230
you're going to give, which I'll tell you very quickly, is OpenAI create.

27
00:01:46,390 --> 00:01:52,630
And you pass in the model name and the messages, and then you're going to print the result.

28
00:01:52,630 --> 00:01:57,830
And the result will be this word response .0. content.

29
00:01:58,270 --> 00:02:02,630
And always look back up in this code and copy and paste to get that example.

30
00:02:02,630 --> 00:02:06,110
And when you've done that you should find that that works.

31
00:02:06,110 --> 00:02:10,470
And if you've built something like something that can give you the subject of an email or something

32
00:02:10,470 --> 00:02:13,990
else, then please consider doing a PR.

33
00:02:14,350 --> 00:02:19,590
The instructions are in the in the guide so that I can get to see it myself.

34
00:02:19,590 --> 00:02:25,350
And I can share in your joy of your first commercial call to an LLM.

35
00:02:25,870 --> 00:02:31,310
And then finally, there's this extra exercise I mentioned for people that are pros, people that already

36
00:02:31,310 --> 00:02:32,230
know this stuff.

37
00:02:32,270 --> 00:02:38,110
You'll be aware that, as I say, my my way of web scraping is super simplistic, but you could use

38
00:02:38,150 --> 00:02:45,070
selenium or playwright or something similar to do proper web scraping with a web page that gets rendered

39
00:02:45,070 --> 00:02:47,870
in the browser, and you'll see that other people have done that.

40
00:02:47,870 --> 00:02:49,590
There's a playwright version there.

41
00:02:49,590 --> 00:02:52,110
I know that there's several selenium versions.

42
00:02:52,110 --> 00:02:54,230
You could look at how other students have done it.

43
00:02:54,230 --> 00:02:59,110
But first, if you know this stuff, if you've used selenium or playwright before, have a shot at doing

44
00:02:59,110 --> 00:03:02,790
it yourself as well to build a more industrial strength.

45
00:03:02,790 --> 00:03:04,390
Web page summarizer.

46
00:03:04,710 --> 00:03:09,870
And then afterwards, as I say one more time, please do consider sharing your code.

47
00:03:10,230 --> 00:03:11,950
The guide will tell you how to do it.

48
00:03:11,950 --> 00:03:13,030
Submit a PR.

49
00:03:13,190 --> 00:03:18,990
I will gladly review it and I'd love to include it in the community contributions with 1 or 2 other

50
00:03:18,990 --> 00:03:20,230
people that have done this too.

51
00:03:20,660 --> 00:03:23,020
Okay, so with that, that's your mission.

52
00:03:23,020 --> 00:03:23,940
Enjoy it.

53
00:03:23,940 --> 00:03:24,820
Fill this in.

54
00:03:25,020 --> 00:03:28,660
If if you if you find that you're struggling with it, please don't worry.

55
00:03:28,700 --> 00:03:33,300
We're going to be doing this so many times over the next eight weeks that's going to become second nature.

56
00:03:33,540 --> 00:03:34,340
Give it a shot.

57
00:03:34,340 --> 00:03:35,420
Give it your best crack.

58
00:03:35,420 --> 00:03:38,940
And if you can't get it this time, don't worry.

59
00:03:38,980 --> 00:03:41,060
We'll be coming back to this tomorrow.

60
00:03:41,300 --> 00:03:41,780
Wow.

61
00:03:42,020 --> 00:03:46,780
You're probably thinking, whoa, this if this is if this day is what every day is like, this course

62
00:03:46,780 --> 00:03:47,980
is going to go on forever.

63
00:03:48,020 --> 00:03:50,500
No, today was a big, big day.

64
00:03:50,500 --> 00:03:52,660
And it's longer than most of our days will be.

65
00:03:52,700 --> 00:03:55,540
You've got through a ton of material.

66
00:03:55,580 --> 00:04:02,220
You've made your first request to an LLM, and I'm hoping that you've also adapted it for a different

67
00:04:02,220 --> 00:04:03,620
commercial task.

68
00:04:03,620 --> 00:04:05,540
And maybe you've even submitted a PR.

69
00:04:05,580 --> 00:04:10,340
So much has got done and and it's just just a great start.

70
00:04:10,380 --> 00:04:13,060
Your environment is built and ready for prime time.

71
00:04:13,060 --> 00:04:16,260
And tomorrow we're going to start getting a little bit deeper.

72
00:04:16,380 --> 00:04:24,340
But already, even though your journey just started, your 2.5% of the way along your journey to the

73
00:04:24,380 --> 00:04:29,420
to the path of being 100% a master of LM engineering.

74
00:04:29,460 --> 00:04:33,060
You've already used a llama to run a model locally on your box.

75
00:04:33,060 --> 00:04:34,180
Perhaps several models.

76
00:04:34,220 --> 00:04:35,660
Seems like an age ago that we did that.

77
00:04:35,660 --> 00:04:36,580
But we did that.

78
00:04:36,860 --> 00:04:41,300
You've written code that calls OpenAI's frontier models.

79
00:04:41,300 --> 00:04:44,380
You know the difference between a system prompt and a user prompt.

80
00:04:44,620 --> 00:04:46,580
And you've built summarization.

81
00:04:46,580 --> 00:04:49,700
You know how to apply to commercial problems.

82
00:04:49,700 --> 00:04:51,820
Maybe you've built a couple yourself.

83
00:04:52,260 --> 00:04:54,660
Tomorrow is going to be a little bit of a breather.

84
00:04:54,780 --> 00:04:56,940
It's going to be an easier day, I promise.

85
00:04:56,980 --> 00:05:01,820
We're going to talk about the steps it takes between now and being a master of LM engineering.

86
00:05:01,900 --> 00:05:03,780
I'm going to set you up for success.

87
00:05:03,980 --> 00:05:09,660
We're going to talk through some of the top frontier models out there, and we'll also do a little bit

88
00:05:09,660 --> 00:05:13,900
of a dig into Olama and using Olama locally.

89
00:05:13,940 --> 00:05:19,900
I kind of repeat of that lab, but using Olama locally on on our own computers, people that didn't

90
00:05:19,900 --> 00:05:21,140
want to to get OpenAI.

91
00:05:21,180 --> 00:05:24,060
That's your chance to use Olama instead.

92
00:05:24,260 --> 00:05:27,060
So I'm super excited to see you on day two.

93
00:05:27,100 --> 00:05:30,620
You survived all of day one that by far the biggest day of the whole course.

94
00:05:31,020 --> 00:05:32,980
So tomorrow much gentler.

95
00:05:33,020 --> 00:05:33,780
Can't wait.

96
00:05:33,820 --> 00:05:34,500
See you then.