Artificial intelligence (AI) technology is rapidly impacting our everyday lives, as people seek to access new tools and shortcuts to boost productivity. The tech is scattered across global news, most recently with the launch of ChatGPT. But what is artificial intelligence and how does artificial intelligence work?
“Machine intelligence is the last invention that humanity will ever need to make.”
Nick Bostrom, Oxford University philosopher
In this explainer we’ll aim to cover those questions, as well as everything else you need to know about AI.
What is AI?
Let’s first make sure we understand what we are talking about. What is artificial intelligence?
According to Google Cloud, who has long been investing in the nascent space (albeit arguably behind its Microsoft competitor) and incidentally recently launched its own AI tool called ‘Bard’, defines artificial intelligence as:
“A field of science concerned with building computers and machines that can reason, learn, and act in such a way that would normally require human intelligence or that involves data whose scale exceeds what humans can analyse.”
Importantly, AI is not a single technology. Rather, it is a term that includes any type of software or hardware component that supports machine learning, computer vision, natural language understanding, natural language generation, natural language processing and robotics.
While it sounds enormously complex (and it sometimes is), there is a simpler way to think about artificial intelligence.
Simply put, AI refers to computer-controlled systems endowed with the intellectual characteristics of humans, such as the ability to reason, discover meaning, generalise and learn from past experience.
And why would we potentially want that sort of capability? The possibilities are endless. Aside from savings in time and costs, AI has the capacity to complete complex and undesirable tasks quicker, easier and more accurately.
From a business perspective, AI offers enormous efficiencies where we’ve seen its main application lie in machine learning and deep learning, used for data analytics, predictions and forecasting and more.
However, let’s not get ahead of ourselves just yet. We’ll dive into specific AI applications later. Before that, let’s talk about how AI works.
How does AI work?
AI machines and systems vary significantly. There is no single standard for how they work as each is designed with a specific purpose in mind, capable of exercising human-like faculties to a lesser or greater extent. Instead, if you want to understand how AI works, it’s best to consider it at a higher level.
AI typically combines large datasets with intelligent processing algorithms to learn from patterns and features in the data it analyses. Much like a human that would take data, interpret it and learn, a machine is doing that for you. Except of course, the machine is exponentially quicker and can process much larger datasets than any human could.
Every time an AI system processes data, it is testing it and measuring its performance based on the metrics being evaluated, and in the process develops additional expertise. Since the AI is never taking a break, it can learn incredibly quickly and accomplish a lot in a very short space of time.
Application of artificial intelligence technology
As noted earlier, there are plenty of different types of AI systems. However, the most important thing to understand is that AI isn’t a single field. There are several components or sub-fields that make up the science of artificial intelligence.
These are the main fields commonly applied by AI technology.
Machine Learning
This application of AI refers to computer systems, programs, or applications that learn automatically and develop better results based on experience, all without being programmed to do so.
It is machine learning that allows AI to identify data patterns and insights, improving the output of whatever task the system is designed to achieve.
Deep Learning
This refers to AI which learns and improves by processing data.
Deep Learning uses artificial neural networks which mimic biological neural networks in the human brain to process information, find connections between the data, and come up with inferences, or results based on positive and negative reinforcement.
Neural Networks
This is a process whereby datasets are analysed repeatedly to find associations and meaning from undefined data.
You can think of neural networks as the network of neurons in the human brain, allowing AI systems to take in large data sets, uncover patterns amongst the data, and answer questions about it.
Cognitive Computing
This is an important aspect of AI systems which allows computers to mimic the way that a human brain works when performing a complex task, like analysing text, speech, or images.
Natural Language Processing
This element of AI allows machines to recognise, analyse, interpret, and truly understand human language, either written or spoken.
This aspect is vital for an AI system that interacts with humans via text or spoken inputs, like Amazon’s Alexa.
Computer Vision
Computer Vision lets AI systems identify components of visual data, like the captchas you’ll find all over the web which learn by asking humans to help them identify cars, crosswalks, bicycles or buses.
What are the 4 Types of AI technology?
As we’ve covered the question of what artificial intelligence is and how it works, we turn to the different types of AI.
At a high-level, artificial intelligence can be categorised into one of four types, each of which is outlined below.
Reactive AI
Reactive artificial intelligence uses algorithms to optimise outputs based on a set of inputs. It is the most basic variety of artificial intelligence as it merely reacts to current scenarios and cannot use taught or recalled data to make decisions in the present. Basically, reactive AI tends to be fairly static, and is unable to learn or adapt to novel situations.
Chess-playing artificial intelligence is a good example of a reactive system. It optimises the best strategy to win the game. Thus, it will produce the same output given identical inputs. Other examples include Netflix’s recommendation engine or spam filters.
Limited memory AI
This variety of AI can adapt to past experiences and update itself based on new observations or data points. However, the extent of ‘updating’ can be quite limited and the length of memory is quite short — hence the name.
One example of this is autonomous vehicles, which can interpret road and traffic conditions and adapt appropriately, even learning from prior experiences.
Theory-of-mind
This type of artificial intelligence technology is fully-adaptive and has an extensive ability to learn and retain past experiences. When machines acquire decision-making capabilities equal to humans, we will have achieved AI theory of mind.
This would include advanced chatbots that could pass the Turing Test, a test to determine whether a machine can demonstrate human intelligence. However, the important aspect here is that the actual machine itself, while intelligent, lacks self-awareness.
Self-aware AI
This is the most advanced form of artificial intelligence. As the name suggests, self aware AI refers to machines or systems that become sentient and aware of their own existence.
It almost sounds like science fiction, but when machines reach this state, they are not only aware of the emotions and mental states of others, but also their own. When self-aware AI is achieved we would have AI that has human-level consciousness and equals human intelligence with the same needs, desires and emotions.
At the moment, this artificial intelligence technology hasn’t been developed successfully yet because we don’t have the hardware or algorithms that will support it.
At present, the majority of frontier work is being done in the limited memory and theory-of-mind AI. The notion of self-aware AI remains a topic of much discussion and debate, ranging from dismissive on the one hand to outright catastrophising on the other.
AI and machine learning
Machine learning is devoted to understanding and building methods that ‘learn by leveraging data to improve performance on a set of tasks. It is viewed as a subsect of artificial intelligence.
Essentially, machine learning algorithms build a model based on sample data to make predictions or decisions without being explicitly programmed to do so.
Machine learning is used in a host of varied applications including medicine, email filtering, agriculture or speech recognition. In its application across business problems, machine learning is also referred to as predictive analytics.
Examples of artificial intelligence technology
As highlighted earlier, artificial intelligence technology is a broad field with applications across countless fields and sectors. Here is a small sample of some of the more interesting uses of AI we’re seeing today.
ChatGPT
ChatGPT, the AI-powered chatbot developed by OpenAI launched in November 2022 and has rapidly risen to prominence in recent months, even integrating with browsers like Bing and Opera.
It uses language models that have been fine tuned to improve the quality of output and the ability to learn. At the core, ChatGPT strives to mimic a human through conversation by asking it questions. However, it is also highly versatile and can be used in innumerable applications including writing software, composing music, fairy tales, and student essays.
As ChatGPT demonstrated an ability to pass various school and university tests, educational institutions are now looking to ban it.
Picture generators
Designers are now having to compete with AI-powered artists as we’ve seen a host of paid and free AI image generation tools proliferate in recent years.
Each has its own features and “templates” of sort, however the gist of it is that you simply need to input the data (perhaps a photo or some text) and click ‘generate’. If you don’t like it, you can change the inputs and generate the art as many times as you like. There’s even a hot debate raging in the art world whether AI art is really art.
Some of the tools are good, others less so. We asked Image Upscaler to draw us a rat in photographic style in a fantasy setting. Here’s what we got.
Navigation maps
This is one of the oldest applications of AI. Google Maps and Waze, amongst others, are constantly updating with live real-time information to help users get from A to B in the most efficient way possible.
From weather conditions through to traffic, these tools are constantly taking in data and ‘learning’ on the go to provide recommendations. They even have enough data so that you can plan your journey in advance, knowing what the ideal time is to leave town to avoid traffic.
Amazon recommendations
Ever wondered how Amazon knows what you are looking for, often before you even start looking for it?
Artificial intelligence technology helps the algorithm understand what you, and people like you, tend to buy. The same principle applies to Netflix. That way, when someone else uses your account, you’ll understand the bizarre recommendations that don’t make sense.
Facial recognition
Most people don’t appreciate that every time you open your iPhone, the device is unlocked using facial recognition software that is powered by AI.
Apple’s FaceID can see in 3D and lights up your face with 30,000 invisible infrared dots and captures an image. It then uses machine learning to compare the scan of your face with what it has stored about your face to determine if the person trying to unlock the phone is you or not. Apple claims the chance of breaking FaceID is one in a million.
What is the biggest challenge facing the government with AI?
It’s not surprising to learn that the private sector has thus far embraced artificial intelligence technology to a much higher degree than governments.
In a report, Deloitte noted that while we were seeing the use of chatbots and algorithms, adoption remained slow due to five key roadblocks:
- Many public sector and government organisations have an elementary understanding of their data.
- Employees often do not possess the necessary AI and data management skills.
- The AI landscape is becoming increasingly complex and competitive.
- There is less encouragement for public sector employees to be innovative and take risks.
- AI algorithms require upkeep from the specific providers, which is an added cost for public sector organisations.
Pros and cons of AI technology
With any technology, there are good and bad elements. Take social media for example – while connecting friends and family around the world, it’s contributed towards higher levels of depression among teenagers.
AI is no different, it is rich in promise but has several notable downsides. First, let’s talk about the benefits. These include:
- No downtime, AI operates 24/7;
- Better, faster decision making by removing bias;
- Fewer errors due to learning capabilities;
- Creates efficiencies and cost savings as “boring” or repetitive tasks can be automated and done by AI;
- Helps lead to the creation of new inventions (such as AI technology that predicts cancer); and
- Can take risks that humans won’t, such as venturing 800 metres below ground in a mine.
Now let’s look at the cons of AI:
- High initial cost, not many individuals or businesses can afford it. Although it inevitably will become cheaper;
- Increases unemployment as those with repetitive jobs are automated away; and
- Lacks human creativity, a certain human quality that isn’t tangible but could otherwise be classified as ‘out of the box’ thinking.
- It can be used for malicious purposes, such as being used to develop weapons or to interfere with elections, a concern Elon Musk expressed last year.
AI and the future: good or bad?
Whether AI is good or bad really depends on who you ask. Most businesses will see the efficiency gains as a good thing. So will the individuals who use things like Google Maps, Uber, iPhones and other consumer applications.
However, some individuals may be concerned as to the extent that automation is disrupting their industry. More specifically, they worry about whether their jobs are at risk. History is littered with examples of people resisting change when their jobs are threatened. Even when the automobile was introduced, many persisted in thinking that horses were better. How things play out in that regard remains unclear.
We also need to think about the role of malicious artificial intelligence technology whether it relates to social engineering, misinformation, hacking or autonomous weapons. Again, it’s probably too soon to make a call on the impact of nefarious AI applications, but it’s worth noting nonetheless.
Ultimately, if we ask ourselves the question of whether AI technology makes our lives better or worse, right now it seems reasonable to conclude that the benefits outweigh the costs. As artificial intelligence becomes increasingly sophisticated, that calculation is likely to change. We’ll leave that for you to decide, sometime in the future.
