15. When LLM alone isn't enough? 👉 APIs for AI?
Models are becoming increasingly sophisticated and massive. Consequently, there are many cases where what was previously provided by code implemented as an agent can now be resolved simply by asking the model a question.
However, despite this, there are still situations where simple calls to the LLM are difficult to resolve.
In this lecture, we will explore how to configure the LLM to execute multi-turns or tool calls, and consider why agents become involved.
00:00 Intro
00:05 Concept of Agent Intervention
00:25 Rapid Evolution of Modern LLM
01:19 Basic Characteristics of LLM - Statelessness and Response to Requests
02:18 Relationship Between AI and Essential Intermediary Services
05:38 3 Reasons Why Agents Are Needed - 1. Statelessness
06:12 3 Reasons Why Agents Are Needed - 2. Hallucination
09:09 3 Reasons Why Agents Are Needed - 3. Differences by Model