Blog icon

AI Explained

It's maths, not magic.

AI is all around us. But how does it actually work?

We believe that understanding AI is crucial for everyone, so we partnered with Google Australia to create this educational video to demystify AI – to show what it is, how it works, and how it's being used right here in Australia. 

Watch our video as we shed some light on what is happening behind the scenes in your favourite AI-driven apps and programs. Scroll down to access resources and see how we're using AI at CSIRO.

[Image shows animated character swiping the air, changing from wearing normal clothes, into scuba diving clothing and then turning into a koala bear.]

It’s behind the scenes in our favourite apps. Making breakthroughs in the lab, and constantly in the news.

[Image changes to an animated cell structure which then changes into news headlines discussing artificial intelligence.]

Artificial Intelligence is all around us.

[Image changes to a male character floating in the air, surrounded by abstract shapes representing artificial intelligence]

But how does it work? Let’s shed some light on the world of AI.

[Image changes to a modern room filled with abstract metallic shapes floating in the middle of the rooms space]

[Image changes to a cluster of different coloured abstract shapes floating in a dark background]

Artificial Intelligence’ is an umbrella term that covers a wide range of systems that all work in slightly different ways.

[Image changes to hundreds of abstract shapes floating together that slowly group and then fly together towards the top of the video frame]

But many of the most widely known use mathematics to find patterns in data.

These patterns are then used to make predictions.

[Image changes to show a sculpture of a cat sitting on a plinth. Many colourful other cat sculptures bounce around the frame in a comical way]

The most common forms of these AI systems use something called machine learning.

Where algorithms analyse data, arranging the patterns and features into models.

[Image changes to show many abstract, bubbly shapes grouped together with individual shapes floating away from the cluster]

[Abstract shapes float away and begin to form groups of similar shapes]

You can think of models a bit like a map.

[Image changes to an animated character unfolding an old-style looking map on a table. We are looking down onto the map]

Let’s look at how a machine learning  model might use an image of a koala.

[Image changes to a tourist looking up to a koala in a tree and then taking a photo of the koala]

[Photo of a koala is shown which then zooms in to the individual and many pixels of the photo]

[Digital pixels then turn into binary code, represented by a series of 0’s and 1’s]

This image has millions of pixels.

[Binary code then morphs into a series of interconnected and colourful lines with horizontal planes intersecting them]

When we put this digital file through a machine learning model, these data points are processed via many layers of multiplication and addition until patterns start to emerge for different features.

[Binary code appears again and floats towards a small metallic looking cube in the centre of the frame. This data-filled cube then floats down onto the previously seen map]

You can think of these like islands.

The more images we add, the more comprehensive the map becomes.

Here on ‘Koala Island’, the west side of the island represents koalas with small ears, and the east, big ears.

[Country shapes of the map begin to morph until they resemble the head and ears of a koala. The words “Koala Island” are written on the map]

What’s really mind blowing, is that ear size is only one feature, and to represent all of the different shapes, colours, moods and compositions of koala images, we not only have to get 3D but imagine thousands of dimensions.

[Photos of slightly different koalas appear on the map which then rise out of the map as 3D shapes. These shapes then shoot into the sky where a koala is seen observing them floating through the sky]

[Camera pans around the sky to show many different representations of koala imagery including drawings through to photographs]

Maybe for now, let’s keep it at just 2!

Now, if we take this map that’s been trained with millions of koala images, and ask what an image might look like… here, an AI system can generate a completely new image related to that location.
This is called Generative AI.

[Small dots fall onto the previously seen map, representing the input of data into an AI model. The AI model then generates a new photo of a koala]

Amazingly, this mapping process works for any data. Whether it’s text, images, sound - whatever can be described with numbers!

[Scene changes to show a series of diverse videos, images, codes and abstract shapes all floating through space together]

When we train models with these different data types together, it’s a bit like combining two maps.
Here we have a text-to-image model, trained using images and their text labels.

[Two maps are seen behind one another floating in space with colourful lines running through both maps, connecting information. Behind the two maps is a series of random images and videos showing nature, animals, landscapes and people.]

These models can take on complex tasks, answer questions, write poems and music, and even generate videos from scratch.

But fundamentally, what we’re doing is giving computer systems a way of mapping information, and making connections between patterns.

It’s maths, not magic.

[A series of tiny spherical green objects are bunched together against a mute-coloured background, resembling the shape of a flower with petals and a stem]

So while these outputs are very convincing, it’s important to understand that they’re  only a prediction based on training data.

[Flower image shrinks to reveal a patchwork of interconnected green square pixelated shapes. Air is pushing through the shapes causing different areas of the image to rise and float as if being caught in the wind]

If we go back to our text-to-image map, and imagine that our text map was trained using American examples, instead of Australian.

[Different American-English words appear on screen in a list. These includes ‘check’, ‘dummy’, ‘flip-flops’, ‘chips’, ‘hood’, ‘trunk’ and finally ‘thong’]

This prompt might give us a much different result.

[Scene changes to show the back of an animated koala wearing thong-like underwear]

Sometimes this can lead to something called bias. Where unfair or unbalanced outputs are generated that amplify inaccuracies or gaps in the data.

[Shot of the same kolas feet, this time wearing ‘thong’ shoes as they are called in Australia]

[Camera zooms out to show the full body of the koala, wearing both the shoes and underwear referred to as ‘thongs’ as well as a scuba mask and breathing tube]

Take planning a playlist for example.

If an AI was trained using only your listening history, it won’t have the right data to generate a playlist that appeals to everyone.

[Three koalas stand next to each other in an Australian bush landscape. The koala in the centre of frame plays a tree branch as you would a guitar]

[Different music types appear as text bubbles above each of the three koalas’ heads]

It’s important to remember that human and artificial intelligence are different things.
For example, humans can instinctively understand context and apply common sense.
AI systems approach things differently.

[Three koalas all dance together]

[Animated character walks through a water-like ocean, standing on an abstract shape floating in the water. More abstract shapes populate the water and sky surrounding the character in the centre of the frame]

That’s one of the reasons it’s important to understand how AI systems work, to ask questions and decide when and how we want to use them.

[Characted walks over the surface of the water while abstract shapes float around him through the sky. Scene changes to show the same character walking through a forest at nighttime. This image then changes to show the same scene in an infrared-style image. Scene changes again to show the character walking through a scientific lab with scans of the human brain appearing on large screens behind them]

In the right hands, AI systems can be incredibly powerful tools.
Helping manage huge data sets, see patterns that humans can’t see and automate complicated processes.

[Animated character is then seen scuba diving underwater in a large giant kelp forest]

But it’s up to all of us to ensure it’s being pointed at the brightest future possible.
To find out more check out csiro.au/ai

[Scene changes to show another animated character walking through an Australian bush landscape, wearing khaki clothes]

[Scene changes to an animated female-appearing character walking in a modern room, surrounding by plants and with a cork board behind her with various photos pinned to it. In the centre of the cork board text reads ‘CSIRO.au/AI’]

[Character exits frame and the camera pans in closer to focus in on the website text on the cork board]

[Video ends]

Transcript

Share & embed this video

Link

https://www.youtube.com/embed/5mHxO2JpCR0?si=eZ3wzgBvYeewydTG

Copied!

Embed code

<iframe src="https://www.youtube-nocookie.com/embed/5mHxO2JpCR0?si=eZ3wzgBvYeewydTG" width="640" height="360" frameborder="0" allow="autoplay; fullscreen" allowfullscreen></iframe>

Copied!

AI at CSIRO

Applying AI to tackle Australia's greatest challenges

For more than 100 years, CSIRO has been using innovative science and technology to tackle national challenges.

Today, we are continuing that tradition by harnessing the power of digital transformation to amplify and accelerate our science, open new frontiers and deliver benefits for Australia. Across the globe, AI is transforming the way we work, socialise and deliver services such as healthcare.

CSIRO is home to one of the largest applied AI capabilities in the world, with more than one thousand researchers working on a diverse range of AI and data science projects.

Our researchers are using AI to tackle challenges from bushfire management to boosting agricultural productivity, improving cybersecurity and protecting our environments including the Great Barrier Reef.

We are working with industry partners to apply AI, for example, helping manufacturers use computer vision to increase production and ensure safety on the factory floor.

We are also developing the frameworks and capability to lead the world in using this powerful set of technologies ethically and responsibly.

/-/media/D61/CSIRO-website-refresh-images/national-cancer-institute-bwMhq_itmMU-unsplash-scaled.jpg?mw=1000&hash=839CA6BD199BA50257ADDBE25D02C9E5

Artificial Intelligence articles

More AI research

Contact us

Find out how we can help you and your business. Get in touch using the form below and our experts will get in contact soon!

CSIRO will handle your personal information in accordance with the Privacy Act 1988 (Cth) and our Privacy Policy.


First name must be filled in

Surname must be filled in

I am representing *

Please choose an option

Please provide a subject for the enquriy

0 / 100

We'll need to know what you want to contact us about so we can give you an answer

0 / 1900

You shouldn't be able to see this field. Please try again and leave the field blank.