OpenAI API

- In a very short time from now, you won't know or in most cases even care what generative AI system you're using. You won't even think of it as using AI. You see, the way we use AI right now as chat bots you have to prompt a very specific way to get them to do roughly what you want them to do is a short detour on our journey to our future with AI. Let me explain. When you check the weather app on your phone do you ever wonder where this data is coming from or do you just assume Apple or Google or Samsung or someone is providing that data for you? Probably the latter, right? Well, what's actually happening is, this app on your phone calls up an API, an application programming interface from a specialized weather data service and requests realtime and predictive weather information for your current location. That API is in the most basic terms a programming interface that lets computers talk to each other without us humans getting in the way. The weather app and all the data transfer that happens invisibly behind the scenes is a good illustration of how specialized data services are integrated into the applications we use via APIs and hints at how generative AI services will be invisibly integrated into pretty much everything we do all through their own APIs. Now, as I record this the main way we interact with generative AI models is through basic chat bots that surface the core AI system. We chat with ChatGPT and it provides responses. We request images from DALL-E and Midjourney and they generate responses. In this direct interchange, we use standalone generative AI apps and they generate output based entirely on their training data and whatever data we provide them with at the moment of interaction. In the immediate future and going forward. Our interactions with generative AI will change from using standalone generative AI apps and services to using them as integrated parts of other services. And in this process, generative AI will go from being sources of information to being intermediaries and interfaces to information. This shift will be facilitated by APIs. To understand this difference, compare ChatGPT with the forthcoming Microsoft 365 Copilot. With ChatGPT, you get a chatbot that outputs coherent human language responses with information of unknown origin. With Copilot, you get contextual help by an AI system based on what you are currently working on, to which you're probably going to say, but Morton, these two are entirely different things, not really. You see, Copilot is powered by GPT, the same GPT you talked to through ChatGPT. But because GPT has been integrated into Copilot through an API and is part of the larger product you don't think of it as chatting with GPT. Instead, you think of it as Copilot. The underlying AI model is relatively irrelevant. That's what these APIs do and that's how they'll change our interactions with the digital world. They make generative AI an integrated and largely invisible part of our interactions. Generative AI APIs provide an interface allowing developers to access the functions and features of API services like GPT and integrate them into their own products and services. They provide tightly controlled and configurable AI on demand, if you will. APIs allow us to seamlessly add AI features into the larger user experience of whatever service or app or platform they're integrated into. And done correctly this integration will be substantively different from chatting directly with a GPT model. First, it likely won't use the core service but instead a hosted secured and provisioned enterprise implementation of the API that protects the privacy and security of the company and its customers. Second, when someone interacts with the AI model embedded in an app the app will often do the prompting behind the scenes and control what types of requests the customer can submit. What data is passed along with the request to the generative AI, what product data sets the AI has access to, and what types of response the AI can return to the customer. Say for example a legal firm wants to add a virtual AI assistance to their website to walk clients through the process of filling out basic legal documents for a lower cost. Instead of building a chatbot from scratch the company can leverage an existing generative AI model like GPT through an API and incorporate it in a chatbot as a service feature on their website. This way, they can offload the responsibility of building and maintaining the AI to the AI service provider so they can focus on building a customer experience around the AI chatbot by incorporating it into their existing services and customizing what the chatbot can do and what type of info it provides to the customer. Like I said at the top, soon you won't know or even care what generative AI systems you're using when you interact with an app just like you don't care what specific weather data collection and prediction service your phone uses to tell you what the weather will be like tomorrow. AI will become a seamlessly integrated part of the apps you use every day if it isn't already. And that's all thanks to generative AI APIs.