As artificial intelligence (AI) and large language models (LLMs) continue to advance, Microsoft is stepping up to support developers in creating their own chatbots. The company recently unveiled Copilot Chat, a sample application based on their open-source Semantic Kernel software development kit (SDK). This SDK provides a foundation for developing custom chatbots with natural language processing and speech recognition capabilities.
Developers can train chatbots using their organization's data through the SDK, which is hosted on Microsoft-owned GitHub. Microsoft's Shannon Monroe highlighted the various advanced features Copilot Chat offers, including multiple conversation topics, speech recognition, file uploading, persistent memory storage, and downloadable bots that can join conversations.
Microsoft's Integration of Generative AI Capabilities
Microsoft has been aggressively integrating generative AI capabilities from OpenAI, such as ChatGPT, GPT-4, and DALL-E, into its product portfolio. In March, the company introduced Copilot, an AI-enabled assistant based on GPT-4, to Microsoft 365. This assistant helps enterprises using Outlook, Word, PowerPoint, Teams, and Excel. Since then, Microsoft expanded Copilot's use into other areas like security products and Visual Basic.
By utilizing Semantic Kernel Copilot Chat, developers can create LLM-based chatbots for various applications, including HR departments, customer service, e-commerce, and education.
Enhancing User Experience with Chatbots
According to Monroe, personalized assistance and natural language processing provided by custom chatbots can significantly improve user experience for customers, students, and employees. Users can easily access information without navigating complex websites or waiting for assistance from customer service representatives.
Chatbots can also boost an organization's efficiency by allowing employees to focus on more complex tasks, generating personalized recommendations, enhancing accessibility, and meeting increasing demand without hiring additional staff.
Getting Started with Copilot Chat
Developers interested in trying Copilot Chat should first update to the latest version of the Semantic Kernel Github Repo to access the most current sample apps. Then, follow the instructions provided to enable optional features like Azure Speech Recognition and Persistent Memory Store. A video tutorial is also available to assist with the installation process.
To run Copilot Chat, developers will need Visual Studio Code, Git, .NET 6.0, Node.js, and Yarn, as mentioned in a support document. The app features a conversation title, a panel displaying various conversation threads, and a conversation thread showing chatbot responses and user prompts. Users can input their prompts in a box at the bottom of the screen.