Lesson video

Hi, I'm Kashif, your computer science teacher for the computer systems unit.

In this lesson, we're going to be learning about artificial intelligence, machine learning, and you're going to be training your very own Google teachable machine.

For this lesson, you're going to need a pen, some paper.

You're also going to need some fruit.

So I've got an apple here, I've got a banana or some props, for example, a Rubik's cube or something that you can train a screen to recognise.

You're also going to need to remove any distractions that are going to get in your way of focusing.

Once you've done that, let's begin.

In this lesson, you will define artificial intelligence and machine learning, discuss moral issues associated with these technologies.

So let's get started with the task.

If you could define artificial.

So an example that could be for example, like flavour, flowers, hair, and so on.

Second question provide synonyms for intelligent.

So a synonym is a different word with the same meaning.

And lastly, when would you call a person, an animal or a machine intelligent? I'll let you write down your answers for that.

Okay, let's check your answers.

So define artificial.

Artificial can be defined as something created by humans, usually a copy or a substitute for something that's natural.

Synonym for intelligent.

Let's see if you got any of these.

So we've got clever, creative, imaginative, ingenious, insightful, inventive, knowledgeable, perceptive, rational, smart, and thinking, inventive like that one.

And lastly, when would you call a person, an animal or a machine intelligent? Now this was a bit of a trick question because it's a tough one to kind of nail down.

I would love to see what you've got for this.

So at the end, I'll show you how you can share your answers with us.

It's interesting that what we might call intelligent for a, a human or a person might not classify as intelligent for an animal.

So the answers are going to be quite interesting to see.

Defining artificial intelligence.

So if you could write a definition of what you think artificial intelligence is.

Right, let's see how you got on.

So it was a tough one and there's no single agreed definition for artificial intelligence.

This is because words like intelligence and thought are difficult to pin down.

And also because machines considered intelligent now will probably be ordinary in a few years.

But if you've got something along the lines of, any machine that performs tasks that typically require intelligence in humans.

Well done, you're along the right lines there.

Okay, so you can see an image here of the legendary HAL 9000 computer from the film, A Space Odyssey, or you can imagine Ultron from the Avengers movies.

Now we're years away from achieving the kind of general artificial intelligence that's portrayed in books and films. Our present artificial intelligence research mostly focuses on individual aspects of intelligent behaviour.

So let's have a look at some tasks and the progress that we've made so far.

So playing ball games like checkers, chess, or Go, the progress that's been made so far, checkers was solved in 2007, computers play perfectly.

Deep Blue by IBM beat the top human player in chess in 1996.

Humans haven't beaten a top chess programme since 2005.

And AlphaGo by DeepMind beat the top human player in Go in 2017.

So it just shows you the progress that's been made in the board games.

So let's have a look at some other tasks, such as proving mathematical propositions and planning and scheduling.

So we've got automated provers have deduced, thousands of known or new propositions, and also discovered shorter proofs.

And in terms of planning, we've got computers are used extensively in manufacturing, crew scheduling, self-driving vehicles and space exploration.

So if you write down, do you think these tasks require thinking by humans? And do you think computers can perform these tasks well? Okay, have a go.

So it's quite interesting, the deep questions to be thinking about, but I'll show you at the end, how you can share your questions and your answers with us.

Let's have a look.

So can the machine do this? AI has succeeded in doing essentially everything that requires thinking, but as failed to do most of what people and animals do without thinking that somehow is much harder.

So we've got some quotations here from books and let's have a look at some other tasks there.

So we've got identifying objects in images, progress so far, the accuracy has jumped from 50% to 90%, since 2011.

And we've got a really fun activity for you to be doing quite shortly.

Next, identify words and sound.

So major advances since 2009.

Error rates have dropped to around 5%.

So if you compare that to professional transcribers, and if you have a look at the applications such as Shazam, you know, they do, they pick up songs and music quite easily.

Generate speech from text.

Major advances in 2016.

So now almost indistinguishable from a real human voice.

How can a machine do this? Handle and manipulate objects.

So the progress so far is that robotic arms that pick up objects is constantly improving.

And this is in the research phase as of 2020.

And walking, two-and-four legged robots are constantly improving.

And they also in the research phase, the AI effect.

So here, we've got some quotes from some articles and some books.

So the first one by Rodney Brooks, every time we figure out a piece of it, it stops being magical.

We say, that's just a computation.

The next one by Nick, once something becomes useful enough and common enough, it's not labelled AI anymore.

And the last one by Douglas, AI is whatever hasn't been done before or done yet.

So here's an example of some artificial intelligence and we've got a route that's being planned and here we've got some of the examples.

So we've got a Google Home and we've got an Echo Show as well.

Can a machine do this? So hold the conversation.

So far, holding an open-ended conversation with a human is considered a benchmark for AI.

No chatbot has recently achieved that goal as of 2020.

And translate between languages.

Major advancements by Google in 2016, systems produce useful output, but still an open problem.

Can a machine do this? So there's three tasks here, understand and answer questions, drive a car and diagnosing medical images.

Have a go and write right answer down, and let's see how we get on.

Okay, let's see how we get on.

So understand and answer questions.

So Watson by IBM beat the top human player on Jeopardy.

So it's capable of providing evidence to justify it's answers to, Drive a car.

It's a highly complex problem there have been some major breakthroughs since 2005, and lastly diagnosing medical images.

So in terms of performance, it's been exceptional and it's been similar to experts since 2012.

Also the recent advances in artificial intelligence are due to breakthroughs in machine learning.

The story so far.

So the goal is to create a machine that performs a specific task.

The method is to programme the machine to perform that task, provide the machine with explicit instructions.

Now for some tasks, providing explicit instructions is far too complicated.

Do you remember that keyword where we've got a sequence of instructions to solve a problem? Yes, it's an algorithm.

So here we're providing algorithms. So some of the tasks are identifying objects and images, identifying words and sounds, generating speech from text and so on.

So how are we going to do it? How are we going to meet our goal? So the method we're going to use is teach the machine to perform the task.

So how do we teach humans? So in terms of teaching machines, how can this be achieved? Have a think about it.

Okay, so we need to provide the machine with the examples.

We need to programme the machine to learn from examples.

So on the left hand side, you can see some pictures of some dogs.

So here were, these are training examples to teach a machine, how to recognise dogs.

This is called supervised learning.

So the next option is training examples to teach a machine, how to play noughts and crosses.

A win results in positive feedback, a loss in negative.

So here we're providing the machine with feedback and the machine is learning from that feedback.

So that's called reinforcement learning.

Machine learning.

For programming computers to learn from experience should eventually eliminate the need for much of this detailed programming effort.

So this is quite interesting because we'll always need programming.

Now, programming the machine to learn and providing it with the necessary training will always be essential.

So machine learning does not eliminate programming.

Task one, be the teacher.

So for this task, you're going to teach the computer the difference between apples and oranges or any of the props or things that you've got home.

You're going to be using a Google teachable machine.

Follow the instructions in your worksheet.

Pause the video to complete your task.

Task one, be the teacher.

Using the worksheet, complete task one.

Resume once you're finished.

How do you get on? So this was actually one of my favourite tasks.

I actually used an apple, I used a banana.

I even used a Rubik's cube as well.

So I really enjoyed this activity.

And I'd love to hear how you got on or what kind of things you used to teach the computer.

Okay, so let's have a look at some moral considerations now.

Thinking beyond the coolness.

So applications of artificial intelligence, we've got our self-driving cars, medical diagnosis, banking, detecting fraud, approving loan and mortgage applications and so on.

So what are the moral considerations when it comes to these? Have a think.

So the moral considerations, when it comes to self-driving cars is what happens when there's an accident, who's responsible? When it's the medical diagnosis, how can decisions be explained? And lastly, when it comes to banking, how can we guarantee that the machine training does not lead to discrimination and how can decisions be explained? Lastly, automation, performing tasks instead of humans.

How will humans handle lower demand for labour? And how will the benefits of AI be fairly distributed? That's the end of this lesson.

Thank you very much for paying attention.

I hope you enjoyed the activities and I hope you got a much better understanding of machine learning and artificial intelligence.

We'd love to hear from you so you can share your work with us.

We've got the Instagram, Facebook or Twitter option that you could tag us with the @OakNational and #LearnWithOak.

Thank you for your time today.

And hopefully I'll see you next lesson.

Take care, bye.