Learn how to build trusted AI applications with large language model APIs, dynamic grounding, prompt templates, and AI-powered orchestration …
Here are highlights from article Building AI-Powered Apps with LLMs and Einstein
1. Building AI-powered apps with LLMs:
– Train your own model:
– Provides full control over the data the model learns from.
– More accurate for specific industry use cases.
– Consider time, resources, expertise, and data security.
– Customize an open-source model:
– Takes less time and is less costly than training from scratch.
– Requires a team of specialized ML and NLP engineers.
– Data security tension may still exist.
– Use existing models through APIs:
– Easiest way to build applications.
– Models not trained on contextual or private company data.
– Output may be too generic.
2. Adding contextual or private company data through the prompt:
– Dynamically created prompts include only data the user has access to.
– Addresses data security tension.
– Techniques to address concerns of passing private data to third-party APIs.
3. Building AI-powered apps using existing models through APIs:
– Basic API call available from major model providers.
– OpenAPI, Anthropic, Google, Hugging Face, and Cohere offer APIs to work with.
You can read it here: https://sfdc.blog/pOTzK
Source from developer(dot)salesforce(dot)com