I give you generative APIs
I love drinking the kool-aid of this new AI age but in the hype cycle we are currently in there is a lot of noise and tools like chatGPT are being overlooked for more simple yet powerful tasks such as acting as an API to the internet.
In this video, I walk through how I iterate on a series of prompts to create an API that itself was written with chatGPT and returns chatGPT generated data in a format (JSON) that can be understood by other systems or to go in a crazy inception loop may be read by chatGPT. It’s all very inception!
You can read through the slides or watch the video below.
The final prompt:
Create a JSON object containing a list of the 10 most populated cities in the world, sorted alphabetically by city name. For each city, include the following properties:
id: an incremental number from 1 to 10
city_names: a list of dictionaries, where each dictionary has the keys language and name, and contains the name of the city in English, French, and German
country_name: the name of the country the city is located in
city_population: the population of the city
description: a one-line sentence describing the city
createdOn: a random date and time in March 2023
imageUrl: a URL pointing to an image of the corresponding country
hash: a MD5 hash of the country_name
The JSON object should have a top-level property called createdOn, which is a string representing the date and time when the JSON object was created.
The api to see the magic:
I would love to hear any feedback/comments on this concept of creating generative AI APIs.
Or if you need any support with your AI strategy please get in touch at leemallon.com.