Microsoft helps devs create chatbots – because who needs human interaction anyway?
It's now easier to whip up the conversational tool of your HR department's dreams
Developers who want to create their own chatbot are getting help from Microsoft, a company itself deeply mired in AI and large language models (LLMs).
Redmond is unveiling Copilot Chat, a sample application based on the vendor's open source Semantic Kernel software development kit (SDK) that can serve as the foundation for developing custom chatbots complete with natural language processing capabilities and speech recognition.
They also can be trained using an organization's own data through the SDK, which is available on Microsoft-owned GitHub.
"With our complete sample, you can take advantage of advanced features such as multiple conversation topics, speech recognition, file uploading to make the chat smarter with your own up-to-date information, persistent memory store that allows the bot to get smarter and smarter with every use, and even downloadable bots to share with others, joining them in the conversation," wrote Microsoft's Shannon Monroe this week.
Microsoft is moving aggressively to integrate generative AI capabilities – the bulk of them from OpenAI including ChatGPT, GPT-4, and DALL-E – throughout its product portfolio. The company announced Copilot, based on GPT-4, in Microsoft 365 in March as an AI-enabled assistant to help enterprises using various services such as Outlook, Word, PowerPoint, Teams, and Excel.
The company has since expanded the use of Copilot in other areas such as with security products in March and Visual Basic in April.
With Semantic Kernel Copilot Chat, developers can create LLM-based chatbots for use in a range of areas, from HR departments and customer service to e-commerce and educational environments.
- Samsung puts ChatGPT back in the box after 'code leak'
- Top Google boffin Hinton quits, warns of AI danger, partly regrets life's work
- Crooks don't need ChatGPT to social-engineer victims, as they're more than happy to demonstrate
- ChatGPT hasn't been around for long and Nvidia already wants to put a leash on it
"By providing personalized assistance and natural language processing, your own chatbot can improve the user experience for customers, students, and employees alike," Monroe wrote. "Users can get the information they need quickly and easily, without having to navigate complex websites or wait for assistance from a customer service representative."
Such chatbots can also improve an organization's efficiency (by freeing employees to do more complex tasks), create personalized recommendations, improve accessibility, and meet growing demand without having to hire more employees, according to Monroe.
Developers interested in trying out Copilot Chat will need to update to the latest copy of the Semantic Kernel Github Repo to ensure they've got the most up-to-date version of the sample apps and then go here for instructions and optional features to enable Azure Speech Recognition and Persistent Memory Store.
There also is a video showing the steps for installing the app. To run Copilot Chat, developers will need Visual Studio Code, Git, .NET 6.0, Node.js, and Yarn, according to a support document.
The app itself includes a conversation title, a conversation panel showing different conversation threads the user is having with the chatbot, and a conversation thread showing the chatbot's responses and the user's prompts. There also is a box at the bottom of the screen where users can type in their prompts. ®