Lesson #4: Compete to train your ML agent

Objectives

In this lesson you will learn:

  • Machines learning agents use data to continue making better decisions.
  • Programmers train machine learning agents to make decisions through practice with lots of data.
  • Machine learning agents trained with different datasets could make different decisions that lead to different actions. This means some agents make better decisions and take more correct actions than others.

To complete this lesson you will need:

  • A device with access to Internet
  • AI Inventors Notebook

Over the course of this lesson you will:

  • Build a AI agent to classify text (train text).
  • Compare actions of machine learning agents trained with different datasets.

Pretend you’re an AI agent

Machine learning and human bias

Activity

Take a marker and sticky note. Take 2 minutes and draw as many pictures of dogs as you can.

You are an AI agents that has been programmer with the goal of finding all the dogs in your neighborhood.

The sticky note drawings are your dataset.

Questions

What patterns do you see in the drawings? Why are you finding patterns? What can you add to your data to help the agent better find dogs?

Machine Learning

AI agents gather data to make decisions that enable actions

Question

How do machine learning agents take actions without humans telling them exactly what to do?

We can train (or program) Machine Learning agents to find patterns in data and continue using data to make even better decisions.

Machine Learning training is a little like teaching your dog a trick.

Let’s Practice : Create an AI Agent

Video tutorial for creating a machine learning agent

You’ve trained a machine learning agent:

  • Cognimates gathers data. You input data like a sensor would collect data.
  • Train agent to practice finding patterns in text data and test if it’s making the right decisions
  • Agent decides if text is positive, negative or neutral
  • That enable actions

Compare Agent’s Actions

Cognimates Image tutorial

Question

Do you think agents that gather different data might make different decisions?

Different decisions can lead to different actions.

Agents with different data make different decisions and take different actions. This means that some AI agents make better decisions and take more correct actions than others.

Reflect

Whole group reflection

Can you think of any AI agents not making the right decisions? How could they be made better?

In your AI Inventors notebook

Do you have an idea for an AI invention that would use sound, text or number data to achieve a goal?

Big Takeaways

Programmers train AI agents to find patterns in data to make decisions – so agents with different data will make different decisions.

An AI agent with not enough data will take the wrong actions. It’ll be hard to find the pattern!