1
00:00:00,000 --> 00:00:10,200
In this video, we'll talk about AI prompt engineering.

2
00:00:10,200 --> 00:00:14,560
Prompt engineering is the process of designing and refining prompts to guide AI systems in

3
00:00:14,560 --> 00:00:18,160
the best possible way, to get efficient output.

4
00:00:18,160 --> 00:00:20,440
And there's a couple of core techniques.

5
00:00:20,440 --> 00:00:22,920
First, be explicit.

6
00:00:22,920 --> 00:00:26,260
The more explicit you are, the better your answer can be.

7
00:00:26,260 --> 00:00:32,779
So not just install a web server, but better use install Apache on Ubuntu 24.04, so that

8
00:00:32,779 --> 00:00:38,139
AI can consider all the differences that might apply if you're not specific.

9
00:00:38,139 --> 00:00:41,900
As for structure, that is also a part of be specific.

10
00:00:41,900 --> 00:00:47,259
If you ask generate a YAML file, generate a configuration file, and so on, then AI

11
00:00:47,259 --> 00:00:50,639
knows what to do, and it can be more specific.

12
00:00:50,639 --> 00:00:53,980
And also, it might be a good idea to provide examples.

13
00:00:53,980 --> 00:00:59,599
If you want a solution to comply to certain standards that are used within your organization,

14
00:00:59,599 --> 00:01:01,959
you might want to feed an example.

15
00:01:01,959 --> 00:01:08,120
So that looks like this, and then you copy-paste the YAML code that you normally would use.

16
00:01:08,120 --> 00:01:09,120
Iterate and refine.

17
00:01:09,120 --> 00:01:14,599
And that means that you need to evaluate the answer that the AI system has given you.

18
00:01:14,599 --> 00:01:19,720
And then you can perfectly ask, hey, improve the previous answer with information about.

19
00:01:19,720 --> 00:01:24,500
And while you do so, make sure that AI doesn't forget about what you did in the previous

20
00:01:24,500 --> 00:01:25,500
answer.

21
00:01:25,500 --> 00:01:29,699
If you are going to iterate, that is a real risk that might be happening.

22
00:01:29,699 --> 00:01:32,980
And also, it's a good idea to break tasks into pieces.

23
00:01:32,980 --> 00:01:37,059
And if you respect all of this, you should be more efficient in your prompt engineering

24
00:01:37,059 --> 00:01:38,300
and get better results.