Introduction
So much have happened when it comes to generative AI lately and a huge part of this is due to the launch of ChatGPT. I have been using it on a daily basis both for private and work related matters. In this post I will share with you how I use it and why it makes me more productive.
ChatGPT privacy concerns
ChatGPT promises not to share your data with anyone else, and it says that OpenAI employees need special permission to look at your conversations. Still, it's really important for you to understand what your workplace expects.
I've requested and obtained a detailed document featuring guidelines such as the prohibition of sharing code that might reveal company secrets. It's wise to sort this out initially, so you're aware of the permissible and impermissible uses of ChatGPT in your work environment.
Communication
When interacting with customers or external parties, I often revise and refine my text. This not only ensures that I don't overlook any details, but it also guarantees that my messages are clear, concise and easy to understand. ChatGPT help me iterate on this more efficiently.
Some commands that I use:
Rewrite the following:Rewrite the following in the [tone]Rewrite the following so that it fit [situation]:
When providing the content, I ensure that any company-specific information is omitted, thereby maintaining a general level of detail that goes in line with the company privacy and ChatGPT policy.
Finally when ChatGPT respond, I look over the content. If it's not good enough, I try add a bit more context to what I think could be improved. When its finally good enough I conduct last manual adjustment and fills in any omitted information.
Boilerplate's
Because ChatGPT cut of date is june 2020 the most up to date solutions might not be possible to find. But some questions are still relevant. Lets say I would like to create a bash script doing the following:
- Retrieve the comma separated list as parameters
- Call the executable in a loop for each element
- Write down the result to a file
This script could take a few minutes to get in place but with ChatGPT you could save those minutes by using the produced boilerplate.
Code changes
This one can be a bit hard to pull off because of policy concerns.. but If your company allows you to put in isolated methods or files that by themselves does not exposes any company secrets, refactoring can be very powerful. Some example commands:
- Convert the following JavaScript file to TypeScript
- Convert the following imperative code to be functional
- Add test scenarios for the following function by using this [framework/tools]
- Add regexp based on the following rules
- Add additional methods for the following code that does [this]
The number of use-cases here is unlimited. The code modifications shared can greatly differ in quality, so my typical strategy is to use them to generate a basic code structure swiftly, and then progressively perfect it.
Other things I sometimes using ChatGPT for
- Documentation. Produce documentation for the following in markdown format
- Summarize the following. In example public customer feedback or an article
- General googling but for ChatGPT
What Im not using ChatGPT for
- Cut off date. Im not using ChatGPT for solutions that requires up-to-date information.
- A lot of context. The max number of tokens aka context length is 4096 tokens. Which makes it hard for ChatGPT to produce useful responses.
- Privacy concerns, both based on the company policies and sensitive personal information.
- Blog posts. I put much effort in my own reflections and thoughts. I might use the rewrite strategy on a paragraph or two but the overall post will be from my own words.
Happy Coding!
