Automating My Job with GPT-3

73 captures<br>26 Jan 2021 - 24 Mar 2026

Dec<br>JAN<br>Feb

28

2020<br>2021<br>2022

success

fail

About this capture

COLLECTED BY

Organization: Archive Team

Formed in 2009, the Archive Team (not to be confused with the archive.org Archive-It Team) is a rogue archivist collective dedicated to saving copies of rapidly dying or deleted websites for the sake of history and digital heritage. The group is 100% composed of volunteers and interested parties, and has expanded into a large amount of related projects for saving online and digital history.

History is littered with hundreds of conflicts over the future of a community, group, location or business that were "resolved" when one of the parties stepped ahead and destroyed what was there. With the original point of contention destroyed, the debates would fall to the wayside. Archive Team believes that by duplicated condemned data, the conversation and debate can continue, as well as the richness and insight gained by keeping the materials. Our projects have ranged in size from a single volunteer downloading the data to a small-but-critical site, to over 100 volunteers stepping forward to acquire terabytes of user-created data to save for future generations.

The main site for Archive Team is at archiveteam.org and contains up to the date information on various projects, manifestos, plans and walkthroughs.

This collection contains the output of many Archive Team projects, both ongoing and completed. Thanks to the generous providing of disk space by the Internet Archive, multi-terabyte datasets can be made available, as well as in use by the Wayback Machine, providing a path back to lost websites and work.

Our collection has grown to the point of having sub-collections for the type of data we acquire. If you are seeking to browse the contents of these collections, the Wayback Machine is the best first stop. Otherwise, you are free to dig into the stacks to see what you may find.

The Archive Team Panic Downloads are full pulldowns of currently extant websites, meant to serve as emergency backups for needed sites that are in danger of closing, or which will be missed dearly if suddenly lost due to hard drive crashes or server failures.

Collection: Archive Team: URLs

TIMESTAMPS

The Wayback Machine - https://web.archive.org/web/20210128080042/https://blog.seekwell.io/gpt3

SeekWell Blog

Automating My Job with GPT-3

Using GPT-3 to generate database-ready SQL to answer business questions<br>As an analyst, I spend a lot of my time writing SQL (or other code) to answer questions about our business. These questions can range from simple customer support queries ("Does user X have the right plan?") to evaluating KPI's and growth metrics ("How many users signed up in the last month and what percent of those converted to paid?") to the more open-ended ("How much revenue will we have in 12 months?").

To make my job easier, I try to automate as many of these questions as I can. My company, SeekWell, builds awesome tools to help with this, like a unified team SQL repository and seamless scheduling of queries, reports, and alerts. Many things, though, like actually writing SQL code, are difficult to automate—or at least have been.

Enter GPT-3<br>Openai's GPT-3 is starting to break the conventional wisdom of what tasks can and can't be automated. At the most basic level, GPT-3 is a text-completion engine, trained on huge swaths of the internet. It takes inputted text and returns the text that it thinks would appear next. Many have already used it to generate HTML and CSS code from specific design instructions. Others have made #1 trending blog posts generated mostly by GPT-3 with some creative prompts.

In my case, since so much of my job is writing SQL, I want to be able to describe a question in plain English and have GPT-3 convert it into the SQL code that, if executed on my Postgres database, would answer the question.

To accomplish this, I found more success using GPT-3 Instruct (still in beta) than the more generalized GPT-3 engines. GPT-3 Instruct (in particular, davinci-instruct) lets you give specific instructions, like "Only respond in correct SQL syntax", that guides GPT-3's responses. If you're interested in trying it out, you can sign up for the waitlist for the GPT-3 API here.

Below, I'll detail my experience with the API. In each example, I'll include a gif of asking GPT-3 a question. Below the gif will be the input question (generated by me, in green) and GPT-3's response translating it into SQL (generated by GPT-3, in blue). I'll also include my instructions to GPT-3 (in yellow) and examples I fed GPT-3 (in orange).

Here's the first instruction I gave it, a question, and the response from GPT-3:

Instruction: Given an input question, respond with syntactically correct PostgreSQL. Be creative but the SQL must be correct.

Input: how many users signed up in the past month?

GPT-3 Response: SELECT COUNT(*) FROM users<br>WHERE...