It’s important for all of us to understand Artificial Intelligence (AI) and get the most out of it.
High-quality data is essential for training AI, especially machine learning (ML) models.
AI not only helps us in our daily lives, but also helps solve big problems like climate change.
Everyone's talking about artificial intelligence (AI): what can AI do, and how can it make our lives easier and more productive?
AI is already all around us – AI helps us when we log into our phone using facial recognition, when we use social media, when we spell check our emails, etc. Is it a problem that most people just think it's OK to use AI without understanding it? We think AI is a big problem. Understanding how AI works can help us determine how it can work best for us.
Here are four ways to help you understand the basics of AI, ask questions, and make informed choices.
AI is a powerful tool with great potential, but it needs to be used responsibly and ethically.
AI is math, not magic
AI is an umbrella term that covers a wide range of systems that work in slightly different ways. Machine learning (ML) is a type of AI that uses models to enable computers to learn from data and perform tasks.
ML uses mathematics to find patterns in data and create models. Creating AI models often requires huge amounts of data. The more data you have to train the model, the better the AI will perform its tasks.
Large-scale language models (LLMs) are a type of AI that recognizes, translates, summarizes, predicts and generates text. They form the core of text-based generative AI algorithms like ChatGPT. They are designed to use and understand language like humans do.
AI creates a world map
ML builds a map of the world that can be used to predict possible outcomes, make decisions, and generate new data.
Check out this video for a visual explanation of how these maps work:
0:00 It's behind the scenes of our favorite apps. 0:02 Achieving groundbreaking results in the laboratory 0:04 And it's always in the news. 0:06 Artificial intelligence is all around us. 0:09 But how does it work? 0:11 Let me explain a bit about the world of AI. 0:15 “Artificial intelligence” is an umbrella term that covers a wide range of systems. 0:19 They all work in slightly different ways. 0:21 But many of the most widely known use mathematics to find patterns in data. 0:27 These patterns are then used to make predictions. 0:30 The most common form of these AI systems uses something called machine learning. 0:36 Algorithms analyze the data and compile patterns and features into models. 0:42 You can think of a model as like a map. 0:45 Let's see how a machine learning model uses an image of a koala. 0:50 This image has millions of pixels. 0:54 By running this digital file through a machine learning model, 0:58 These data points are then processed through many layers of multiplications and additions. 1:02 Until patterns of different characteristics begin to emerge. 1:06 You can think of these as islands. 1:09 The more images you add, the more comprehensive the map will be. 1:13 Here on “Koala Island,” the western side of the island represents a koala with small ears. 1:19 And in the East there are big ears. 1:22 What's really surprising is that ear size is just one feature. 1:27 Expressing a variety of shapes, colors, moods and compositions 1:31 In the case of koala images, we not only obtain 3D images 1:35 But imagine thousands of dimensions. 1:39 For now, let's just stick with two! 1:43 Now, take a look at this map, which was trained on millions of images of koalas. 1:47 Now, let's ask what the image looks like… 1:51 The AI system can generate entirely new images relevant to that location. 1:56 This is called generative AI. 1:59 Amazingly, this mapping process works for any data. 2:03 Text, images, sounds, anything that can be expressed with numbers! 2:08 When you train a model using these different data types together, it's like combining two maps. 2:14 Here we have a text to image model trained using images and their text labels. 2:21 These models perform complex tasks, answer questions, and 2:25 You can also write poetry, music, and generate videos from scratch. 2:31 But fundamentally what we're doing is giving computer systems a way to map information. 2:37 Create connections between patterns. 2:40 It's math, not magic. 2:43 These results are quite compelling, 2:46 It's important to understand that these are just predictions based on training data. 2:50 Returning to the text to image map, 2:53 Now imagine that TextMap was trained using American examples instead of Australian examples. 2:59 This prompt can lead to very different results. 3:03 Sometimes this can lead to what is called bias. 3:07 It produces outputs that amplify inaccuracies and gaps in the data. 3:12 For example, plan your playlist. 3:14 If the AI is trained only on your listening history, 3:18 It doesn’t give you the right data to generate a playlist that will appeal to everyone. 3:22 Humans and artificial intelligence 3:25 It's different. 3:26 For example, humans can instinctively understand context and apply common sense. 3:32 AI systems approach things differently. 3:35 This is one of the reasons why understanding how AI systems work is important. 3:40 Ask questions and decide how and when to use them. 3:44 AI systems can be incredibly powerful tools if used properly. 3:49 Helps manage huge data sets and find patterns invisible to humans 3:54 Automate complex processes. 3:57 But we all have a responsibility to ensure we are looking towards the brightest future possible. 4:02 Visit csiro.au/ai for more information.
The output of an AI may be impressive, but it is only a prediction based on the data used to train the model. If you train a model on the wrong data, you will get useless results.
For example, in February 2024, a passenger sued a major airline after the airline's chatbot gave them incorrect advice regarding a refund request. It turned out that the model the chatbot was using had been trained on outdated data.
AI can predict future scenarios, enhance decision-making, and reveal new insights.
AI can’t do everything
Despite some dystopian futures depicted in movies and TV shows, AI will not take over the world. AI will not be more “human-like” than humans anytime soon.
This is because human intelligence and artificial intelligence are different.
Humans are able to instinctively understand situations and make decisions based on common sense. Humans are able to exercise judgment, empathy and creativity. Humans are able to recognize the past, present and future and the emotions that accompany each era.
So far, AI has been based on patterns in the data it was trained on, and as a result, there are significant limitations to what it can produce.
Some researchers say calling AI “intelligent” is misleading because all it really does is pattern matching. “Intelligence,” as we commonly understand it, is much more than that.
Three teammates gathered around the robotic arm.
What's next for AI?
When used properly, AI systems are incredibly powerful tools, able to recognise patterns invisible to humans and automate complex processes.
AI has a huge role to play in solving our nation's biggest challenges, like climate change and the energy transition, but it's up to all of us to make sure it's used for the right reasons and at the right time.
Understanding the basics of AI and how it works can help you make informed decisions about when to use it in your own life and can also help educate others.
