Natural Language Tasks

Superpower to Gain

With Natural Language Tasks, you can talk to a computer in your own words and receive totally human-sounding text in response.

Target Output

The goal for this module is to familiarize yourself with OpenAI’s GPT-3 tool. The output you will end with is a Prompt structure that you have tested and refined through which GPT-3 analyzes any transcript text that you provide it. You will use the transcript text from the last modules as an input for this work.

Steps to Take

Click below to sign up for OpenAI
Navigate to the Playground interface where we will begin to interface with the GPT-3 service
Try to type in a variety of different commands into the Playground interface to become familiar with how it works. As a simple example, if you’re just looking for a place to start, try asking it to write a poem about your hometown.
Pick one of the sample Podcast Transcripts from the list below, and click to open the Google Sheets link. Each row of the file represents one speaker’s segment of the podcast. Pick one segment, copy the text and paste it into the GPT-3 playground
Add text before the pasted transcript that says “Podcast Transcript: “. Then at the bottom of the transcript add a new line and insert “Question: Does this podcast transcript mention any books?”
👉
We said in the demo above that we would provide a few other sample transcribed podcast snippets that you could use to test. Here they are: He bought me, and I'm sure you read these books, too, the Art of Hitting 300 by Ted Williams. I believe the Charlie Loud book. And then he bought me The Mental Game of Baseball by Carl Keel and Harvey Dorfman. For whatever reason, bro, I fell in love with that mental game of Baseball with Harvey Dorfman. Here we go, everybody. And what I started to do was in my free periods in high school, I'd go and I'd highlight. Did you really? Incredible. And I'm not kidding when I say this I read that book. This frivolous luxury is the way it was really thought of as how attractive you were to the person. A different idea started to become more common. In the 17 and 18 hundreds, Jane Austen, the famous novelist, may well have been the trailblazer. For those who don't remember the plot of her book Pride and Prejudice. Mr Darcy's, who has been promised in marriage to his wealthy cousin, falls instead for Elizabeth Bennett, a woman of modest means that throws his aunt into a rage. Mr Darcy is engaged. Like first, the COO said the pipeline project, supported by the infrastructure bills, quote, gaining momentum, and projects are moving through the sale cycle into delivery. End quote. In other words, that huge slug of federal spending is finally starting to come through. I like that. At the same time, Jacobs also does a lot of design and engineering for advanced semiconductor plants. Hey, that isn't exactly what the government has with that heavily subsidized chips act. Unfortunately, semiconductor. She also has been a serious downturn. Right now, however, Jacobs says that they haven't seen.
Try asking different questions about the transcript in order to continue to build familiarity with how GPT-3 works
In order to refine your prompt and improve the likelihood of getting good results, try adding a couple line breaks after “Podcast Transcript” and after your question to make it more readable for GPT-3
Then, on the right side of the page, familiarize yourself with the controls that are available. You can experiment, but ultimately reduce the Temperature to “0”, reduce the Maximum length to “4”, and reduce Top P to 0.2
Finally, change the question to ask specifically for books mentioned in the podcast transcript

Next Module

API Calls

Intro Video

NLTs-202212

160.21 MB