How Developers Can Use GPT-3 in Their Products — 3 Real World Tools
Examples to use GPT-3 in interesting ways
A few weeks ago I talked to a bunch of my developer friends about GPT-3 use cases. They all love GPT-3, but I was surprised to find that they think it’s mostly used for marketing applications.
Of course, I don’t agree with them. Although the most talked about GPT-3 use cases are in marketing, there are some really interesting ways to use GPT-3 to help developers.
In this article, I cover 3 tools that make use of GPT-3 in very interesting ways. They are 100% non-marketing — they are designed for developers, DevOps engineers, and SREs.
Hopefully, after reading my article you will have some working ideas on how to apply GPT-3 to the projects you are working on.
Let’s dive in!
Zebrium
Zebrium has been using its own machine learning and combining it with GPT-3 for quite some time, and it is the most mature tool on the list.
About the tool
Zebrium was named a 2020 Gartner Cool Vendor and it made it to the Gartner’s Top 25 Enterprise Software Startups To Watch In 2020.
What is Zebrium?
Zebrium is a log anomaly detection tool. It uses unsupervised machine learning (ML) on logs to automatically find the root cause of software incidents in the original files and then, it delivers simple summaries.
This is where the benefit of GPT-3 comes in. The summary of the root cause is provided in plain English.
What the product does
Zebrium applies unsupervised machine learning to analyze the massive volume of logs generated by software. It identifies and reports the root cause of the incidents without any manual work.
Root cause reports are delivered in the form of a short list of log events. Typically, anywhere from 5 to 20 events. Then, the GPT-3 engine translates the technical details of the log content into a short summary that explains the problem in a way that people who do not have experience with logs can understand.
How does it use GPT-3?
Apart from its primary goal of decreasing troubleshooting time by having ML identify key event sequences, Zebrium also worked on presenting the results in a readable and concise form.
The list of the key event sequences is passed to the GPT-3 engine with an appropriate prompt.
The result is a short, easy-to-understand plain language summary of the problem. This is how Zebrium creates root cause analysis (RCA) reports that can be comprehended by anyone.
Practical Examples
Let’s have a look at a few examples of using GPT-3 by Zebrium.
Example one: Database shutdown
The test was initiated by the administrator shutting down a Postgres database backing up the Atlassian software stack. It caused a stream of numerous errors in logs.
Zebrium’s ML generated a root cause report containing a small set of log lines as well as the following GPT-3 description: “The database server was stopped by the administrator”. This simple sentence was enough to immediately explain the nature of the problem.
Example two: Out of memory
In this case, Zebrium’s ML pinpointed a set of a dozen log lines that showed a server had run into an out-of-memory condition.
The log lines were passed to GPT-3 which generated the description as follows: “The Memcached process was killed by the OOM killer because it was consuming too much memory”.
Who does it help
Zebrium’s technology is designed for SREs, DevOps engineers, developers, and technical managers. It’s also useful for writing up root cause post mortem reports.
Project status
Zebrium is a fast-growing startup. It already has dozens of customers who use the solution to reduce the Mean-Time-To-Resolve software incidents.
GPT-3 plain language summarization is a standard feature of the Zebrium product.
How to test
- Go to their website and press the big blue button “Get Started Free”. Next, enter your data, and receive a free trial.
- Deploy a log collector on your app or use Zebrium’s test app.
- Break the app yourself or just wait for a real issue. You will be pleasantly surprised with the results. You’ll get Machine Learning generated RCA reports and a summary in plain language generated by GPT-3.
SeekWell
SeekWell helps to write SQL and synchronize the results to the other apps used in a company.
About the tool
Being an analytics and data visualization tool, SeekWell acts as a link between databases, such as Postgres, Snowflake, Redshift, and MySQL, and common apps such as Google Sheets, Excel, Slack, and email.
What does the product do?
SeekWell uses SQL requests to connect databases with different applications. It helps synchronize data that allows team members to work in alignment with each other, ensuring an efficient flow of the data.
How does it use GPT-3?
It uses GPT-3 Instruct to convert plain English into SQL requests for syncing work. By giving specific instructions, e.g. “Only respond in correct SQL syntax”, SeekWell helps users without knowledge of SQL to get the data they need from the database.
The tool supports MySQL, Redshift, MS SQL Server, Postgres, and Snowflake. SeekWell works with SQL requests.
Who will benefit?
SeekWell can be useful for developers, product managers, analysts, and anyone who needs to sync databases with supported applications.
Project status
In general, the tool is fully functional, but the GPT-3 Instruct component is still in beta.
How to test
If you want to test SeekWell you can request a trial or demo on the main page. However, for testing GPT-3 functions, you should sign up for the Open AI API Waitlist.
HelpHub
The next tool is called HelpHub, this is a simple and modern knowledge base.
About the tool
HelpHub hosts a fully-featured knowledge base with a friendly and minimalistic design. This tool provides analytics for the articles in the knowledge base and collects user feedback.
In general, HelpHub is designed to maintain a well-structured knowledge base. Additionally, one of its main features is that it can be used to create support articles in a few seconds!
What does the product do?
It is a handy tool for building and maintaining a fully-featured knowledge base. HelpHub uses Artificial Intelligence (AI), and GPT-3 as the technology behind the feature of creating technical support articles.
How does it use GPT-3?
Specifically for the articles, HelpHub applies technology called the AI Article Writer powered by OpenAI’s GPT-3 API.
Who will benefit?
Creating support articles in seconds frees up time for developers, product managers, and technical support. By using HelpHub, fewer people are involved in putting together comprehensive content to help their users.
Project status
The project is already available but seems to be keeping a low profile.
How to test
- You can find a request for a free trial on the official website. Enter your data and sign up.
- Create a section. It can be treated as a folder for the articles based on the topic.
- Create an article and publish it to see the result.
Conclusion
In conclusion, GPT-3 has a larger sphere of application than thought initially.
This technology can definitely simplify routine tasks for developers, SREs, and DevOps engineers.
Though many tools are still testing the possibilities of GPT-3, we have already found some good examples of its successful application.
I hope you enjoyed this article and will consider using GPT-3 in one of your next projects.
Have you tried any other tool with GPT-3?
If so, please share your thoughts and experiences. I would love to hear from you.
