Unlock the Potential of Open AI in Smart Manufacturing
Figure 3-9 : Copilot conceptual diagram 11 . The framework consists of the key components below :
• The Natural Language Interface allows users to express their intents and queries in plain English and receive responses and suggestions from the AI system .
• Large Language Models : o
A code generation engine that uses OpenAI Codex model to generate code snippets or scripts based on the user ’ s natural language input and execute them on the target platform or device .
o A data engine that uses OpenAI Codex ( LLM ) model to perform data processing , visualization , and reasoning over various data sources such as sensors , logs , databases , etc .
• A knowledge graph that stores and retrieves relevant information and best practices for manufacturing processes , systems , and products .
• The AI Orchestrator , which co-ordinates with all other components to execute the complete flow .
For an input natural language question , it first calls into Embeddings and Similarity models 12 to understand the keywords in the question . Based on those it can decide the information retrieval systems it should use – example if it a simple facts-based question it can search the answer in an index and / or it can use the embeddings to filter the contextualized request-response pairs and perform prompt engineering 13 for the LLM .
11 https :// learn . microsoft . com / en-us / microsoft-365-copilot / microsoft-365-copilot-overview
12 https :// openai . com / blog / introducing-text-and-code-embeddings
13 https :// www . promptingguide . ai / techniques 60
March 2024