Discover Some Challenges of Artificial Intelligence

In the past few videos, you learned about the benefits of AI and how people use it every day.

In this video, you'll learn about some of its challenges.

These challenges include AI: making mistakes, showing bias, and automating human jobs.

First, AI can make mistakes.

It's programmed by humans, who might make mistakes when they program it.

And since AI learns by looking at data, it can make mistakes if it doesn't get enough data.

This is why it's often important for humans to work with AI, so they can check what it's doing.

Google Translate is an AI tool that learns from hundreds of millions of already-translated documents on the internet.

But the translations aren't always perfect.

For example, Google Translate may have difficulty translating idioms.

"Idioms" are expressions that have a different meaning than the literal definition of the words in the expression.

An example is "it's raining cats and dogs," which means that it's raining heavily, not that cats and dogs are actually coming down from the sky.

If the AI translates the sentence word for word it is not going to make sense in a different language, but with enough examples it can learn a correct translation.

To help Google Translate avoid these and other mistakes, people can become contributors and verify, or check, translations.

If many contributors verify a translation, it gets a badge.

This is an example of how many people can work together with an AI system to make it more accurate.

Another challenge with AI is bias.

"Bias" means to prefer one person or thing over another.

Consider the following situation where AI might learn bias from data.

Imagine a company is hiring a lot of people.

They get so many applications that they get an AI system to help them decide who to interview.

The AI system is given data about who the company has hired in the past.

Without realizing it, the company previously hired mostly younger people.

The AI learns to imitate this pattern and favors younger candidates over older candidates.

This is an example of how AI can learn bias from decisions made by humans.

AI can also be used for other decisions.

For example, banks might use it to decide who gets a loan for a house.

Or schools and teachers can use it to grade student assignments by programming it to look for specific keywords.

These AI systems can return a grade within seconds.

In each of these instances, AI makes decisions quickly and saves employers, banks, and teachers time.

But can you think of any possible risks of using AI in these ways?

How might these risks be minimized?

Another challenge is the long-term effects of AI automation.

Machines that use AI can do some tasks that used to be done by humans.

For example, some factory work is now done by machines.

And self-checkout machines now do some of the work of grocery clerks.

What might happen as a result?

For example, if robots around the world do more work, might humans get more time to relax?

If AI does repetitive tasks, can people end up with more interesting jobs?

And although AI may automate some jobs that are now done by humans, do you think it may also create new kinds of jobs?

What kinds of jobs do you think might be needed?

Many people and organizations are thinking hard about these challenges, and ways to ensure AI is used responsibly.

For example, Google has shared principles for building AI that is good for society, unbiased, and safe.

In the next video, you'll start creating your project.

Before that, think about what you have learned so far about how AI is used and its benefits and challenges.

Spend a few minutes thinking about this question: What did you learn so far about AI that you never knew?

Go back to your document and type at least three new things you learned.

Now, it's your turn: Think about what you learned so far about AI that you never knew, and type at least three new things in your document.