Intended audience: Social Impact Professionals
The final step in getting started with Generative AI is to set up tools and integrate them into your technology stack to give proper access to everyone. This article is the bridge between the broad strategic elements of this Playbook and the technical implementation, diving into implementation specifics for technical audiences.
To start to implement GenAI at your organization, use tools developed by others or integrated into your existing systems (ChatGPT, Copilot, etc.). Then, begin to develop and integrate your own tools, if your organization has the capacity and knowledge for ethical application. What this looks like will vary based on the organizational need and use case selected.
Here’s one example of what this looks like for a social impact professional first getting started with an idea for using Generative AI:
- Try to set up a basic chat completion endpoint in code and try a few different prompts (see the next section below for instructions).
- Maybe use ChatGPT’s paid subscription service to see how OpenAI does question-answering when you provide a pdf as a knowledge base (or test the equivalent from a different provider, or even multiple vendors).
- Focus on an intuitive grasp on generative question and answer capabilities before diving into developing a proof-of-concept, if you’re considering developing an AI assistant.
- If in-house technical expertise is low, it would be more beneficial to start by evaluating Generative AI requirements using easy to set-up codebases.
- Then eventually, develop and integrate into your own tools, if your organization has the capacity and knowledge for ethical application.
- Later, once the requirements are clearly understood and you can articulate the specific needs and goals, lean on a more technical person to set up the proof-of-concept.
Setting up GenAI chat endpoints: A quickstart guide
Here is a sample step-by-step guide to complete #1 above. Note that this guide is NOT meant to be used for any experimentation that will involve private data. Please only use this for testing with interactions that do not contain proprietary information or anything that should be kept secure.
Alternatively, this OpenAI on Azure service allows for a more private (although also more expensive) use of OpenAI’s tool. Ensure you are mindful of OpenAI’s data privacy in all that you do and make decisions accordingly, even in your initial tests.
Please see other “scoping” articles for social impact organizations, just like seen at the bottom of the page here: https://datakind.github.io/scoping.html
Contributer(s): John Harnisher, Matt Harris, Lindsey Washburn, Kathleen Yaworsky
Contact us
If you would like to learn more about us, partner with us, or get in touch, email us at community@datakind.org
Subscribe