Capabilities

[00:00:00]
>> Yeah, we have a large prompt and we are going to pay for the prompt every time. That's how GPT works. It's part of the deal. Okay, cool. Let me just save that for later uses just in case. Okay, let's go back here. So remember as you see there, prompt is king here.

[00:00:23]
Everything is about the prompt. Capabilities that you can do with the prompt, ask for summaries, inferring things like ask for the sentiment, for example, or relevant data or tags. So let me show you one example of that. So, give me, so with the sentence in triple quotes, infer if the sentence is positive or not, answering with the words or with the letters P or N, let's say, or negative, here, without any other character.

[00:01:16]
And now I can pass a quote, like '''I hate this course.''' So, Send. N, negative. '''I like this course,''' P. So that's inferring, and you can use that to check what the user is typing, what the user is saying. And even look at this, instead of that, select or pick one emotion related to the phrase.

[00:01:53]
Okay, so I like this course. Let's see the emotion. It's that one. I hate this course. That's emoji, by the way. It rains outside. That's emoji. Can you see the emoji there? So this is pretty good for doing these things. Right, so these are different ideas that you can use with prompting.

[00:02:19]
And by the way, this is probably one token. So that's a cheap prompt, at least in the answer. Okay, so that's inferring, actually. You can also transform data, translations, format conversion. Extract data, like you have a text and try to extract the numbers or the emails, or the amount of wherever that you expect the text you have.

[00:02:46]
Create content, expand on the fact, so you have a fact and you can explain, well, write me an article as if you are in the New York Times or in the style of Jack Spear about blah and blah. Okay, so that's prompt engineering.