AI systems already make real decisions, what you see on social media, whether an application gets a second look, what price you’re shown online. In this challenge, you’ll investigate how those decisions actually get made.
Step 1
We are learning to
compare how humans and AI learn to recognise things.
A human can often learn from surprisingly few examples. Machine-learning systems usually need far more data to recognise useful patterns reliably.
I can
identify at least one genuine difference between human learning and machine learning.
Step 2
We are learning to
understand why the examples an AI learns from shape what it can do.
Machine learning works by finding patterns in examples. Change the examples and you can change what the system learns.
I can
explain why the data used to train an AI affects what it becomes good — and bad — at doing.
Step 3
We are learning to
connect training data to something a real AI system can actually do.
Pick the level of investigation that works for you.
I can
give a real or well-reasoned example connecting training data to an AI capability.
Step 4
We are learning to
understand how limited, unbalanced or incorrect data can produce unreliable AI behaviour.
AI does not automatically know when its experience of the world is incomplete.
I can
explain how a weakness in training data can become a weakness in an AI system.
Step 5
We are learning to
combine evidence and reasoning into one clear explanation for another person.
Your finished piece should answer three things: How does AI learn? How is that different from human learning? Why does the training data matter?
I can
create a clear finished explanation showing how AI learns from data, how that differs from human learning and why the quality of the data matters.
Step 6
We are learning to
connect what we now know about AI learning to how we use and judge AI ourselves.
ReflectWhat surprised you most about how AI actually learns?
ApplyHow might knowing about training data change the way you judge an AI answer or output?
Look ForwardWhat would you now want to know about the data behind an AI system before trusting it?
I can
use what I’ve learned about training data to ask better questions about AI systems.
