 Next, we've got nonprofit expert Jason Shim, who's going to be talking about AI from a different direction around using AI text generation for nonprofits. So I'm going to be talking today about AI text generation and I'm going to jump straight into it. You may have heard of a transformer model and you'll be hearing more about them in 2023. And what it was is a model that was developed by the Google Brain team in their AI department and it was open sourced in 2017. And what came out of that were a couple of subsequent models namely BERT and GPT, which I'm going to be spending the bulk of this presentation talking about now a GPT stands for generative pre-trained transformer. And what does pre-trained mean? It means that it was trained on huge bodies of text on the internet, including Wikipedia, as well as something called the common crawl, which is petabytes of data crawled from across the internet. And this is hosted by an organization called OpenAI. Now OpenAI, one of their significant investors is Microsoft and Microsoft owns and GitHub co-pilot was something that came out in 2021 and what it is essentially, you can input comments about what you would like a snippet of code to do and it will autocomplete the rest. Now, it may not always be accurate or correct, but it's co-pilot. So it's alongside a developer that is coding this up and it can significantly accelerate the software development process. Now, why am I sharing all this? And why is this relevant for nonprofits? Similar technology now exists for a general text completion. And so chat GPT is the latest release in a series of GPT models. There has been GPT 2, 3 and chat GPT is considered 3.5. And here's a quick example of how chat GPT could be used with a prompt as a creative experiment to write me a poem about technology in a non-profit sector. And it generated this in a few seconds, a technology of forestill grand in the hands of a nonprofit, a helping hand. It's not the best poem in the world, but it did generate this in a few seconds, take it for the time being requires some editing. But in terms of a practical application, something to keep in mind is it could be potentially used for generating drafts. So let's say if you're writing a grant and here's a pretty standard question that you'll find in a grant, describe your project in detail, including proposed timeline and specifically how you use the funds. And here's the response. The actual response is about five paragraphs long by no means what I say that you should do copy and pastes directly to a grant application, but this could be used to help scaffold a potential response and to help think through how might you integrate some content that is particular to your organization to supplement what is here. Another example would be around like things like marketing drafts. So here's a prompt where it's draft a blog post about how GPT can be used to help nonprofits. And this, the blog post that was generated was about five paragraphs long, you can make it longer or shorter. And again, like the blog post itself, it's okay, it's generic, but it's more around generating that first draft for potential use. So this is something to keep in mind as something that has potential in this area. Now one thing to keep in mind too is that it can be confident, but not necessarily correct. And you definitely a human in the loop there. So here's an example of that confidence that's not necessarily correct. The question how many nonprofits are there in Canada is that it's saying that there's 170 registered charities, 170,000 registered charities in Canada. Now that number is charities and nonprofits number, but it's stating it here quite confidently. And the actual number is closer to 86,000. It does require fact checking and that this really is just pulling from the scouring the internet, but it's not necessarily correct all the time. So what did nonprofits need to keep in mind has a potential use case, first drafts, it could be used and let's say you need to put together something, whether it be a draft of a document or potentially you need to draft an email, you can integrate certain elements where it's okay, can you help draft an email that is going out to a donor also indicate that you just saw them recently, I gala asked them about the following things as well, and it can draft up something in a few seconds. Again, you wouldn't send it outside unseen, you would still need to review it. And on that point, human review is important when you're ever working with AI at this point, making sure that have people to ensure that all the facts are correct in it, and that it's still reflecting your organizational tone, and that it's more of a supplement in that regard. And keeping in mind the quality inputs are needed for effective outputs. In addition to the pre train models, it's expected that you should in the future be able to input additional texts. So let's say if you have a case for support, which may make a great piece of text for a model to to look at as well as things like brand books for your organization. So that indicate how do we talk about organization? What is the tone that you use these kinds of things would be helpful for informing future prompts that you would use for AI text generation. It's not going away anytime soon. This was a tweet last night from Satya Nadella, the CEO of Microsoft and that chat GPT is coming to the Azure Open AI service, so it will be integrated into some of the broader offerings that Microsoft is offering in the future. And moving into the future, there are other models on the horizon GPT for which is going to be the next iteration of the GPT models as well as Lambda, which is Google's offering and others are on the horizon. My suggestion to everyone here is I try it out. I'll drop a link in the in the chat for chat GPT as well if you haven't tried it already. And one thing to keep in mind is that we this is a very quickly evolving space and that the would be helpful for folks to find their own use cases for it. Once again, if you want to ensure that the future with AI is going to be useful for your organization, jump right in there and also experiment with it a little bit and figure out what the use case may be for your particular organization or use case. If you'd like to stay in touch, also drop the more information in the chat to connect on LinkedIn.