1
00:00:00,270 --> 00:00:02,820
-: So before we get into the actual prompting,

2
00:00:02,820 --> 00:00:05,880
let's first talk about prompt priming.

3
00:00:05,880 --> 00:00:08,520
Now, prompt priming refers to the practice

4
00:00:08,520 --> 00:00:11,850
of providing some initial input to the model

5
00:00:11,850 --> 00:00:13,500
before generating your response.

6
00:00:13,500 --> 00:00:16,800
You're basically training it on a specific dataset.

7
00:00:16,800 --> 00:00:20,430
So this input is intended to guide

8
00:00:20,430 --> 00:00:23,280
a model like ChatGPT, towards generating a response

9
00:00:23,280 --> 00:00:25,440
that's more relevant, coherent,

10
00:00:25,440 --> 00:00:29,100
and trained for the user's intended output.

11
00:00:29,100 --> 00:00:32,490
And you can use entire pieces of text

12
00:00:32,490 --> 00:00:33,960
that you've written in the past.

13
00:00:33,960 --> 00:00:35,820
So you could literally paste in.

14
00:00:35,820 --> 00:00:38,580
If you are a writer, you could paste in all

15
00:00:38,580 --> 00:00:41,010
of your past work from the last few months.

16
00:00:41,010 --> 00:00:43,830
And ChatGPT will essentially understand how

17
00:00:43,830 --> 00:00:47,670
to replicate your speaking style, tone, grammar,

18
00:00:47,670 --> 00:00:49,470
and the words that you use.

19
00:00:49,470 --> 00:00:52,080
And you can use that with any type of data.

20
00:00:52,080 --> 00:00:55,050
So let's dive into these examples that further explain

21
00:00:55,050 --> 00:00:57,420
how prompt priming works.

22
00:00:57,420 --> 00:01:01,023
So for our first example, without priming our prompt,

23
00:01:01,023 --> 00:01:04,410
our question is, where should I go on my next vacation?

24
00:01:04,410 --> 00:01:08,550
So ChatGPT has no context, no data

25
00:01:08,550 --> 00:01:11,610
or details on how to effectively answer our questions.

26
00:01:11,610 --> 00:01:15,630
So it's gonna just tell me any destinations around the world

27
00:01:15,630 --> 00:01:17,610
with no criteria.

28
00:01:17,610 --> 00:01:20,550
Now, if we go in and we prime our prompt,

29
00:01:20,550 --> 00:01:22,770
we can say something like, I like to go on a trip

30
00:01:22,770 --> 00:01:24,300
with my wife and kids.

31
00:01:24,300 --> 00:01:26,430
The location should be tropical.

32
00:01:26,430 --> 00:01:28,290
We'd love to be close to the beach,

33
00:01:28,290 --> 00:01:31,230
and I'd like the flight to be a direct flight to

34
00:01:31,230 --> 00:01:36,030
and from LAX, and my travel budget is $5,000.

35
00:01:36,030 --> 00:01:37,410
Then I ask my question,

36
00:01:37,410 --> 00:01:39,750
where should I go on my next vacation?

37
00:01:39,750 --> 00:01:42,000
And the priming output is one suggestion

38
00:01:42,000 --> 00:01:43,950
for your tropical family vacation

39
00:01:43,950 --> 00:01:48,300
within a $5,000 budget can be Cancun, Mexico

40
00:01:48,300 --> 00:01:50,400
direct flights from LAX are available,

41
00:01:50,400 --> 00:01:53,136
and you can find affordable all-inclusive packages

42
00:01:53,136 --> 00:01:56,100
with pricing, starting around a thousand per person

43
00:01:56,100 --> 00:01:57,570
for a week long stay

44
00:01:57,570 --> 00:02:01,050
and popular family activities you can do in Cancun include

45
00:02:01,050 --> 00:02:02,850
snorkeling, visiting ancient ruins,

46
00:02:02,850 --> 00:02:05,160
and exploring local beaches and restaurants.

47
00:02:05,160 --> 00:02:08,460
So it addressed who I was with, where I wanted to go,

48
00:02:08,460 --> 00:02:09,810
my budget that I had,

49
00:02:09,810 --> 00:02:13,920
and having a direct flight from where I live.

50
00:02:13,920 --> 00:02:17,310
Now, the next example without priming the question was,

51
00:02:17,310 --> 00:02:19,770
please create three potential titles

52
00:02:19,770 --> 00:02:23,340
for my new online course that teaches how to use AI.

53
00:02:23,340 --> 00:02:25,530
Now I'm gonna get an extremely generic response,

54
00:02:25,530 --> 00:02:29,520
but what I did instead to craft a better output was this,

55
00:02:29,520 --> 00:02:31,590
please create three potential titles

56
00:02:31,590 --> 00:02:34,680
for my new online course that teaches people how to use AI.

57
00:02:34,680 --> 00:02:38,610
Here's an example of my recent course titles,

58
00:02:38,610 --> 00:02:41,910
please emulate the style and written format of these.

59
00:02:41,910 --> 00:02:44,940
So we have one of my past courses video editing masterclass,

60
00:02:44,940 --> 00:02:47,970
edit your videos like a pro, cinematography masterclass,

61
00:02:47,970 --> 00:02:49,680
the complete videography guide.

62
00:02:49,680 --> 00:02:51,960
So it looks like I used two past examples.

63
00:02:51,960 --> 00:02:55,350
And here were the outputs, AI mastery, the Complete Guide

64
00:02:55,350 --> 00:02:58,950
to Using Artificial Intelligence, the AI expert.

65
00:02:58,950 --> 00:03:02,310
Learn how to or learn to harness the power

66
00:03:02,310 --> 00:03:04,230
of artificial intelligence.

67
00:03:04,230 --> 00:03:06,742
And then AI foundations, a step-by-step guide

68
00:03:06,742 --> 00:03:09,750
to using artificial intelligence.

69
00:03:09,750 --> 00:03:11,940
So now you should understand how important

70
00:03:11,940 --> 00:03:14,760
and effective prompt priming is.

71
00:03:14,760 --> 00:03:17,370
So whether you're training the model on pages

72
00:03:17,370 --> 00:03:21,030
of your past writing style, or just giving some more context

73
00:03:21,030 --> 00:03:23,400
and guidelines for ChatGPT to follow,

74
00:03:23,400 --> 00:03:25,590
this is gonna help you really start

75
00:03:25,590 --> 00:03:29,343
to use prompt engineering and get more effective outputs.