WEBVTT 00:01.260 --> 00:03.690 Instructor: In this video, we're going to understand 00:03.690 --> 00:06.270 the components comprising of a prompt, 00:06.270 --> 00:10.830 and we're going to formally define the terminology 00:10.830 --> 00:13.323 of a prompt and its components. 00:14.640 --> 00:18.300 Former terminology helps us establish a common language when 00:18.300 --> 00:19.860 we talk about AI. 00:19.860 --> 00:22.860 This shared understanding facilitates collaboration 00:22.860 --> 00:24.570 and makes it super easy for us 00:24.570 --> 00:27.180 to communicate ideas regarding AI. 00:27.180 --> 00:30.660 So just like in chemistry where we have the terminology 00:30.660 --> 00:33.330 or in math that we have different terminology, 00:33.330 --> 00:36.960 the same thing will go to AI and to prompt engineering 00:36.960 --> 00:39.180 after we'll formally define what is a prompt 00:39.180 --> 00:40.380 and what it's made of, 00:40.380 --> 00:45.030 it will help us customize our prompts and to optimize it 00:45.030 --> 00:46.950 because we'll understand the elements 00:46.950 --> 00:48.750 that the prompt is comprised of, 00:48.750 --> 00:51.720 and we'll know how to identify and what to address 00:51.720 --> 00:54.963 and which part of our prompt would need modification. 00:56.430 --> 00:59.400 So let's start by describing what's a prompt 00:59.400 --> 01:02.430 when we talk about AI language models? 01:02.430 --> 01:07.050 A prompt is the input that we give to the AI model 01:07.050 --> 01:10.320 that we wanted to produce an output. 01:10.320 --> 01:14.070 You can think of a prompt as a guide for the model, 01:14.070 --> 01:16.710 helping it understand the context, 01:16.710 --> 01:20.070 to process the information, and to generate a relevant 01:20.070 --> 01:22.653 and meaningful response that we can use. 01:23.730 --> 01:25.530 We're going to break the prompt into 01:25.530 --> 01:27.690 four different components. 01:27.690 --> 01:31.320 So let's now describe what are the components of a prompt. 01:31.320 --> 01:34.110 The first component is the instruction, 01:34.110 --> 01:37.800 and you can think about it as the heart of the prompt. 01:37.800 --> 01:42.090 The instruction tells the AI model what task 01:42.090 --> 01:43.590 it needs to perform. 01:43.590 --> 01:45.780 Whether you're looking for a text summary, 01:45.780 --> 01:48.540 or translation, or classification, 01:48.540 --> 01:52.713 the instruction sets the stage for the AI's response. 01:54.390 --> 01:56.400 Next is the context. 01:56.400 --> 01:59.400 So the context provides additional information 01:59.400 --> 02:02.970 that helps the AI model to better understand the task 02:02.970 --> 02:05.730 and to generate more accurate responses. 02:05.730 --> 02:09.450 So for some tasks, context might not be necessary. 02:09.450 --> 02:12.540 While for other tasks, it can significantly improve 02:12.540 --> 02:14.073 the AI's performance. 02:15.630 --> 02:18.720 The next component is the input data. 02:18.720 --> 02:20.850 So it's the information 02:20.850 --> 02:23.700 that the AI model will process in order 02:23.700 --> 02:26.220 to complete the task that we've set. 02:26.220 --> 02:28.590 It could be a piece of text, an image, 02:28.590 --> 02:31.353 or any data relevant for the task. 02:32.850 --> 02:34.980 Now, the last component comprising 02:34.980 --> 02:37.443 of a prompt is the output indicator. 02:38.670 --> 02:43.350 It signals the AI model that we expect a response now. 02:43.350 --> 02:46.470 So sometimes the output indicator is implicit in the 02:46.470 --> 02:49.710 instruction, but sometimes it's explicit 02:49.710 --> 02:52.443 and we'll see example of it very, very soon.