ChatGPT, Generative AI, & AI Alternative Use Cases for Your Nonprofit (Special Edition)

This one is dedicated to nonprofits and nonprofiteers. This is a special edition because I want to give actual use cases that I think you will find valuable for your approach to exploring and implementing AI into your nonprofit. First, though, let's ensure we are all on the same wavelength regarding definitions and foundational understanding.

No. This is not entirely about ChatGPT - it just starts there. This is about bridging the gap between using ChatGPT as an individual and leveraging tools that keep your data secure and benefit your entire organization internally and externally.

AI Definitions and Foundational Understanding

AI: AI stands for artificial intelligence. AI refers to machines or software that can think and make decisions like humans, helping us (humans) solve problems and perform tasks more efficiently.

Generative AI: Generative AI is a culmination of multiple technologies put together. However, it can generate new content, such as images, text, or music, and is inspired by the patterns it has learned from existing data. There is a knowledge plateau with most if not all forms of generative AI. Generative AI has many applications and can even drive functionality when applied correctly.

Machine Learning (ML): Machine learning is like teaching a computer to learn from examples. It's how we make computers smart without explicitly programming every detail.

Deep Learning: Deep learning is a subset of machine learning where artificial neural networks mimic the human brain to solve complex problems, like image recognition or language understanding.

Neural Network: A neural network is like a virtual brain made of interconnected nodes (neurons) that work together to process information. It's used in AI and machine learning to learn from data and make predictions, mimicking how our brain processes information through interconnected neurons.

Large Language Model (LLM): A large language model is like a super-smart computer program trained on massive amounts of text from the internet. It can understand and generate human language, making it capable of tasks like answering questions, writing essays, or even having a conversation with you. It learns patterns and rules from all the text it has seen, allowing it to mimic human language abilities on a large scale.

GPT: GPT stands for Generative Pre-trained Transformer. This model architecture was created by OpenAI and jumpstarted the wave of AI advancement that we are all a part of right now.

Natural Language Processing (NLP): NLP is the AI's ability to understand, interpret, and generate human language, making it possible for machines to converse with us.

ChatGPT: ChatGPT is an application (it is not AI itself). It is a conversational user interface that actually interfaces with large language models (LLMs) known as GPT-3.5-Turbo and GPT-4 (effectively the AI).

Long-Term Memory (LTM): LTM is a part of AI's memory where it stores important information for later. It's like your computer's hard drive for knowledge. LTM is not something like the last 4 chat pairs referenced in a prompt sequence but rather more like data stored permanently. Think of it like the human brain's ability to memorize important things or observations.

Supervised Learning: In supervised learning, AI learns from labeled examples. It's like teaching a dog tricks with treats as rewards – the AI learns from the correct answers.

Unsupervised Learning: Unsupervised learning is when AI learns from unlabeled data, trying to find hidden patterns. It's like discovering new recipes from a pile of ingredients without labels.

Reinforcement Learning: Reinforcement learning is like training a pet. The AI learns by taking actions and getting rewards or punishments, gradually getting better at a task.

Reinforcement Learning from Human Feedback (RLHF): RLHF is an approach within the field of reinforcement learning. RLHF aims to make reinforcement learning more adaptable to real-world scenarios where human expertise is essential for training and fine-tuning agents, such as in recommendation systems or robotics.

Okay, now that that is out of the way...

AI Assistant: ChatGPT is an AI assistant. Tools and apps built for companies that are built to produce specific generative outputs are assistants. This is a loose term that covers things like virtual chat assistants, community outreach assistants, appeal assistants, copywriting assistants, etc.

AI Agent: ChatGPT is not an AI agent. AI agents are typically determined by their level of automation or procedural workflow execution. AI agents run autonomously, complete tasks, do things like search the web or execute functions and do so with the freedom to think and act on their own until the desired goal is reached.

How Does Generative AI Work

First, let me break down exactly how ChatGPT works so that you can see just how things are firing. It is amazing to me even articles that illustrate how transformers operate have mentioned that ChatGPT is a large language model. No. No. No.

ChatGPT is NOT a large language model.

ChatGPT is an application that interfaces with a Large Language Model (LLM), specifically based on the GPT (Generative Pre-trained Transformer) architecture. It is not the LLM itself but rather a user-friendly interface or system built around the underlying GPT model.

Here's how it works:

1. Large Language Model (LLM): The core of ChatGPT is a powerful LLM, such as GPT-3. This LLM has been trained on massive amounts of text data and can understand and generate human-like text.
2. ChatGPT Interface: ChatGPT is designed to provide a user-friendly way to interact with the LLM. It takes user inputs, processes them, sends them to the LLM, and then presents the LLM's responses to the user in a conversational format.
3. Enhancements and Controls: ChatGPT often includes enhancements and controls to make interactions more useful and safe. For example, it has features for controlling the tone of responses through instruction, turning off chat history/data store, and options for adding plugins and beta features for GPT-4 (ChatGPT Plus users). It now also has an enterprise-level plan.

To help you visualize how ChatGPT works by serving chat history to the LLM here is a diagram we made when we were building our own chat history feature for some of our apps. What I would note is that the chat history has a roll-off and you have likely experienced this when ChatGPT seemingly has forgotten someone that it once knew because you told it.

Here we have the following:

1. A system prompt that is present at all times
2. A greeting - ChatGPT does not have a greeting but most applications like to greet the user when they first land on the page for instance.
3. A user prompt - this is what the user has just typed in
4. An assistant response - this is the response from the LLM
5. ^ Repeat steps 1 (a constant), 3, & 4

In summary, ChatGPT is a user-facing application or system that leverages the capabilities of a Large Language Model to provide human-like conversation. It serves as an interface to make it easier for people to interact with and benefit from the underlying LLM's language understanding and generation abilities.

ChatGPT and Nonprofits

ChatGPT is for individuals. I can't speak to the Enterprise plan that ChatGPT now has but even if I could, my thoughts will still apply without a doubt. I have a perfect example for nonprofits that I think will make sense. Let's say that we have a volunteer program where a digital orientation makes sense.

Here is what I WOULD NEVER tell a volunteer:

• Go to ChatGPT
• Sign-in and select GPT-4
• Go to the list of GPT-4 plugins
• Download the [Insert org name here] ChatGPT Plugin
• Insert these custom instructions (potentially paragraphs worth of text)
• Now tell it you are ready for orientation.

Here is what I WOULD tell a volunteer:

• Scan this QR code (or visit this URL)
• Click on the green button to start your orientation

(An AI-powered conversational application walks them through orientation at their pace)

Both applications will effectively use a chat interface to orient the volunteer. The last example just uses a custom-built application that interfaces with a LLM and does not use a public-facing application. This conversational interface is one that could be embedded on a page on your website. Behind the scenes, it would be driven by generative AI in that we would instruct the model to follow a step process and guide the user through orientation step-by-step conversationally.

Think of this like building a website. When we built custom websites for our clients at my digital agency, we would typically use WordPress. WordPress is the most popular content management system (CMS) out there. When we were done, you could not tell from the front end that the site was built with WordPress. We would build a custom theme that would leverage WordPress and benefit from its baked-in functions but we would add additional functionality and stylizations on top.

We are entering a similar era for AI. Specifically, generative AI powered by LLMs. This is what led me and my team at Bold Crow AI to build the "WordPress for generative AI" in our new product called Corvus. I won't go into full detail but you could also think of it like Wix, Squarespace, or the like but for generative AI applications.

This is all about building creative user experiences powered by generative AI and ensuring that it is accomplished responsibly. Remember, generative AI has the power to enhance not only things like content but also operations, data management, and more. When done correctly, the result is tons of time saved, smoother operations, better data and reporting, and ultimately happier nonprofiteers because their workload becomes easier and they can spend time doing the things that truly matter like building relationships, spending quality time with quality data, etc.

Ways for Nonprofits to Creatively and Responsibly Leverage Generative AI

Before I get into specific use cases, one place to start is by asking your team members how they are currently using ChatGPT. Make sure that organizational data is not flowing into ChatGPT via a copy/paste job by a team member using ChatGPT.

After that, see how you may be able to better the experience if you were to build or commission an app that accomplishes what they are currently doing. This would be a way for you to also enhance what they are doing by ensuring that organizational policies, strategies, content guidelines, brand guidelines, and the like are in place and baked into the application. This will lead to more reliable and desired outputs by the AI-powered app.

Okay, I would now like to give you some ideas as to how generative AI can be used by your nonprofit. Let's start with my example above.

1. Volunteer Orientation Assistant

An AI-powered reliable and on-demand volunteer orientation assistant. This might be for volunteers in the field who have a question or volunteers showing up in the morning who want a walk-through of their day. I am thinking about camp staff and volunteers in situations where they may be spread out. This app can have features baked into it that help with a step process and a conversational component as well. Similarly, it can have a data input field where these volunteers can also collect information in the field that they are walked through based on where they are in the process. It would be like having an expert on their shoulder helping them every step of the way and would greatly cut down on incomplete data.

2. Translation/Interpreter (Multicultural) Assistant

In real-time, an AI assistant can translate multiple languages which would instantly break down barriers. In the same way as the volunteer orientation assistant, this app would benefit from having a fine-tuned model or configurations that instruct the model to respond in a certain "organizationally approved" way. This comes down to the betterment of user experiences and ensuring that translations from one person to another are, in a sense, reviewed before being translated by the organization but in real-time against a dataset/training materials.

3. Community Outreach Assistant

We all know that AI can produce content and a lot of it at scale. However, these content pieces may or may not be aware of your organization, and current guidelines, your up-to-date mission, and know what biases typically plague your organization's generated content. With a flavor of reputation management, a community outreach assistant can help humans be more proactive and dial in on what the community is most receptive to at any given time. This comes down to data analysis, guidelines similar to the ones mentioned above, and any other flavors that influence how the assistant responds or handles certain types of requests.

4. Advocacy Coordinator Assistant

In the realm of nonprofit advocacy, an AI-powered assistant ensures your organization's advocacy efforts are well-informed and strategically impactful. It analyzes policy documents, tracks trends, and crafts persuasive messages, that align with your mission and guidelines. This assistant streamlines campaign logistics tracks performance, and enhances engagement, ultimately driving more effective advocacy efforts. Additionally, there could be an agent component to this that at certain intervals would automate gathering content, distilling that content appropriately, and generating the desired content or notifications.

5. Educational Training Assistant

This assistant harnesses the power of AI to enhance learning experiences. It personalizes online training, providing real-time tutoring and tailored learning paths. This assistant supports learners, explains complex topics, and ensures educational content aligns with your organization's goals. With its assistance, nonprofit educational programs become more engaging and effective. As well, the number of questions is drastically decreased because there is now a conversational interface. Similarly, less time is spent searching for an answer that may or may not ever be found.

BONUS!

Data/Reporting Assistant

This assistant/agent is a reliable partner to everyone in the organization and makes data accessible. It automates data analysis, and reporting tasks, saving time and ensuring accuracy. It removes the need for SQL queries to be written against a database and enables anyone to ask questions in their natural language. This assistant generates insightful reports, tracks key performance indicators, and aids in evidence-based decision-making. Your organization can harness data and become a data-guided organization even if your data today is relatively unstructured, scattered, or considered incomplete.

AI Agents & Assistants for Nonprofits

Really, any of the "assistants" can have "agent" functionality baked into them where the agent portion of the AI-powered application is triggered by something and then automatically begins working in the background. The beauty is that generative AI enables all of this to happen with fractions of the amount of development work that was needed previously. While this can impact the cost of development and reduce the turnaround time, nailing the right configurations and the right baked-in prompts for ensuring reliable functionality is paramount.

Complex workflows and automation take careful consideration but when done right, the results are hugely beneficial and the ROI is off the charts. This is not always about the things in your day-to-day that you wish you never had to do again (although this can factor in). Think creatively, think about ways to streamline operations, think about ways to ensure consistency and reliability, and most importantly think about ways to make people's lives and work better.

AI & Nonprofits in Summary

People hear the phrase "AI" and in some cases, they think that AI will replace their job. It will likely replace aspects and certain work they do but in all cases, what I suggest is that AI is here to augment and help make jobs easier. Not that we need to date ourselves but think about when your nonprofit first realized that having a quality website would be the most beneficial thing you could spend marketing dollars on that year. You evolved as a nonprofit and the nonprofiteers working there evolved in their day-to-day tasks. Maybe fewer brochures were designed, maybe less time was spent pounding the pavement for awareness, and it is likely that more time was spent on things like SEO, graphic design, and the like.

As we evolve with AI and march into this future frontier, think about ways to relieve burnout, think about ways to defeat the overhead myth within your organization, and think about ways to creatively approach your organization's operations. I believe this is an opportunity for nonprofits to take back the reigns, use data they were not able to before, retain more donors by providing better experiences, and provide the utmost that they can to their mission.

------- ?? -------

Hey, I'm Dave! I'm a former digital agency owner, now co-founder of Bold Crow AI. I help businesses and organizations implement customized AI solutions in a responsible way. I've built cool tools for nonprofits too, helping them gather and leverage social proof.

Let's connect:?Dave Norris

#futurefrontier #ai #nonprofits