From the course: AI Evaluations: Foundations and Practical Examples

Unlock this course with a free trial

Join today to access over 25,200 courses taught by industry experts.

Why a lot of AI agents fail

Why a lot of AI agents fail

From the course: AI Evaluations: Foundations and Practical Examples

Start my 1-month free trial Buy for my team

Why a lot of AI agents fail

“

- [Instructor] Before we go too far in believing that these agents can do everything we human learned to do in millions of years of learning, let's acknowledge the fact that AI agent do fail. And let's see three main reasons why these AI agent fail. The first reason is we chose the wrong problem. This is not the problem that needs to be solved by AI agents. And that could be either it's a simple problem that can be solved by rule-based system or it's too complex and has no patterns for AI agents to learn to solve that problem. For an example, for the last 10 years, I'm hearing that we want to build an AI agent that does stock transactions for me and always make profit. It's just hard to build an AI agent because the stocks movement is not based on past patterns. It's a very complex dynamics and that changes and learnings from the past cannot be used to assure you consistent returns. And if you try to build that agent, it will fail. The second one is bad taste. A poor product design or…

Contents