Titlebook: Building Generative AI-Powered Apps; A Hands-on Guide for Aarushi Kansal Book 2024 Aarushi Kansal 2024 Artificial Intelligence.Generative A

旁觀者

Introduction to Generative AI,and at some point, a manager is probably going to ask you “can we do generative AI too?” or you’re going to get tempted and hack together an LLM-powered bot at 2 a.m. This chapter introduces you, a software engineer, to the booming world of AI, by cutting through all the hype and demystifying AI. I

Cpr951

LangChain: Your Swiss Army Knife,apter introduces you to LangChain, your Swiss Army knife to building robust applications on top of LLMs and other models. As you build applications beyond just making API calls, you’re going to need various components to connect a model to your own data, to external data, and services, and that’s wh

多產(chǎn)子

Chains, Tools and Agents,bot that answered your questions . could remember the rest of your conversation. This allowed the LLM to become “smarter” by getting context from history. Your chatbot also had access to up-to-date, personal information via a vector database, meaning it was able to answer questions beyond what it wa

規(guī)范要多

Guardrails and AI: Building Safe + Controllable Apps, your day for you. This agent was able to reason and have access to “the world” via API integrations (the so-called tools). This was a fairly simple application, but it was still autonomous . and when AI is autonomous, there’s always space for things to go wrong if proper safeguards are not in place

發(fā)微光

Finetuning: The Theory,rdrails around ensuring your LLM stays on topic, executes the right flow, and is able to block users. You looked into NeMo and understood how it combines LLMs, Colang, and embedding models to create a generalized set of rules, based on natural language rules you give it.

BILK

Finetuning: Hands on,n models. You learned about the whys, whats, and hows of fine-tuning. You learned that fine-tuning can be less resource and time consuming than building and training a model from scratch. The previous chapter talked to you about what happens to the neural network during the fine-tuning process . spe

invulnerable

背書

Outspoken

GLOOM

https://doi.org/10.1007/978-3-540-24785-2ory. Your chatbot also had access to up-to-date, personal information via a vector database, meaning it was able to answer questions beyond what it was trained on. This also helped prevent hallucination.