WEBVTT

00:00.000 --> 00:01.906
-: Hey, and in this video,

00:01.906 --> 00:03.656
you're gonna learn about the least to most technique

00:03.656 --> 00:05.310
and how you can utilize that inside of ChatGPT.

00:05.310 --> 00:06.900
Well, what is least to most?

00:06.900 --> 00:09.390
Basically, you don't always have to have all of the context

00:09.390 --> 00:13.080
inside of a large language model to get started with a task.

00:13.080 --> 00:15.160
Maybe just start with a couple of bits of context

00:15.160 --> 00:17.130
and then iterate on the output

00:17.130 --> 00:20.790
that you get back from ChatGPT or Claude or Gemini.

00:20.790 --> 00:22.830
And then once you've reached a final output

00:22.830 --> 00:25.710
that you're happy with, then you can consider the task done.

00:25.710 --> 00:27.390
Let's have a look at an example of this.

00:27.390 --> 00:30.810
For example, if we were creating an article outline about

00:30.810 --> 00:33.960
what type of food supplements I should take, we could say,

00:33.960 --> 00:37.320
I want to create an article outline

00:37.320 --> 00:40.840
about different types of food supplements

00:41.730 --> 00:43.410
that I want to take.

00:43.410 --> 00:45.930
And there's not that much context within this prompt,

00:45.930 --> 00:47.520
and because of that, you are going to get back

00:47.520 --> 00:48.990
quite a vague response.

00:48.990 --> 00:51.030
We want to give it more context in the introduction

00:51.030 --> 00:52.620
so we can select this section

00:52.620 --> 00:55.372
and then we can say, improve the introduction

00:55.372 --> 01:00.372
and make sure to include habits, frequency, et cetera.

01:02.730 --> 01:06.900
Expand this section, but keep the rest the same.

01:06.900 --> 01:09.420
And so the idea is you're basically taking

01:09.420 --> 01:11.340
a very vague response

01:11.340 --> 01:13.380
when you didn't have a lot of original context,

01:13.380 --> 01:16.650
and as you see ChatGPT generate a response,

01:16.650 --> 01:18.960
you are iterating through the output

01:18.960 --> 01:20.760
and you'll basically keep doing this

01:20.760 --> 01:23.340
until you figure out exactly what kind of format

01:23.340 --> 01:24.300
you want for the output.

01:24.300 --> 01:26.220
The benefit of doing this means that you don't have

01:26.220 --> 01:28.050
to have a very large prompt at the start.

01:28.050 --> 01:30.390
You can work iteratively with ChatGPT,

01:30.390 --> 01:32.100
and you can get a refined output

01:32.100 --> 01:34.920
by selectively saying this needs to change

01:34.920 --> 01:36.150
or that needs to change.

01:36.150 --> 01:38.340
In the next video, we'll have a look at some best practices

01:38.340 --> 01:39.240
for how you can make sure

01:39.240 --> 01:41.493
you get better outputs from ChatGPT.