So… What Really Is AI?

When you think AI, short for artificial intelligence, you might think futuristic robots, sentient machines, and metallic humanoids that clean your home and babysit the kids. Scary cool. These exciting and once only imaginable possibilities are soon to become our surreal realities. Thankfully, we’ve still got some time to wrap our heads around what these exciting and unnerving developments mean for our society.

From sci-fi movies to university studies, the topic of AI has been in our zeitgeist for decades now; but it hasn’t demanded space the way that it does today. In 2023, we saw the rise of its interest, discussion, and even its application across several industries. In 2024, no matter where we turn it seems we can’t escape it. AI this, AI that – no matter where I turn it’s “Ay ay ayy.” With its use and demand seemingly skyrocketing overnight, it leads many of us to question – So…

What Really Is AI?

Artificial intelligence is technology that enables computers and machines to simulate human intelligence, creativity, and problem-solving capabilities necessary to perform tasks. These tasks include recognizing speech, decision-making, and identifying shapes and patterns. The self-training of AI models is known as Machine Learning; “the subset of AI that enables computers to learn from training data… using its group of algorithms designed to automatically learn and improve from experience.” Basically you sit baby AI down to read the hugestest encyclopedia until they’re grown and smart as a big boy.

Once each model of AI is trained on its desired algorithms and data, they are considered a large language model (LLM). Therefore, AI chatbots such as Google’s Gemini, Bing’s Copilot, and OpenAI’s ChatGPT are considered LLMs.

Large Language Models

We’re teaching and training AI large language models to perceive and process information as humans do. In certain aspects, they’ve already surpassed our processing capabilities due to their ability to process tons of data and multiple possibilities at once. With a prompt in the form of a question or request, chatbots (LLM user interfaces) can generate an adequate, unbiased, and unopinionated response.

The level of AI where LLMs can limitedly operate based on our programming or input is referred to as artificial narrow intelligence (ANI). Examples of these are voice assistants or a car’s AI self-driving capabilities. Many AI researchers and trainers are pushing for AI to reach artificial general intelligence (AGI, also known as Strong AI). AGI refers to the state of being where AI machines have become self-aware, possessing an understanding of the world, others, themselves, and can even plan for the future. If they were to evolve past AGI to the point that they become self-aware and more intelligent than their human counterpart, they’d earn the level of artificial super intelligence (ASI). Basically, they’d be metallic sentient beings. Scary… cool?

The potential of AI stands to revolutionize almost every working industry. It’s already increasing our knowledge and efficiency and will undoubtedly be replacing the need for human employees at the helms of most tools and machines; at least in the warehouse environment. Because of this, we must ask ourselves, will this usher us towards a utopian society, or is this the beginning of a dystopian tragedy? Read on to learn more about this new and shiny prospect, its beneficial application in our modern world, and of the risks we take to reach these new technological heights. Firstly, to better understand our new proposed helpers, we must identify the four (main) types of AI.

The 4 Types of AI

1. Reactive Machines

Reactive machines, the most basic type of artificial intelligence, only “react” or offer an output based on the input received. As such, they don’t store new information nor do they use previous observations to inform future decisions. They can only perform certain predefined tasks based on their machine learning. This means that reactive machines operate on artificial narrow intelligence. Examples of this sort of AI would be IBM’s chess-playing program Big Blue, spam filters, or Netflix’s recommendation engine.

2. Limited Memory Machines

Machines with limited memory possess a limited understanding of past events. They can interact more with the world around them than reactive machines can. “However, machines with limited memory cannot form a complete understanding of the world because their recall of past events is limited and only used in a narrow band of time.” This means that limited memory machines operate on artificial narrow intelligence.

For example, some of Elon Musk’s Tesla automobiles are able to drive themselves using their AI technology, ‘Tesla Vision.’ Tesla Vision uses the auto’s cameras and sensors to recognize and identify vehicles and objects within their surroundings. Using the real-time information, it can make turns, adjust speed accordingly, and even brake if need be. Sounds like a technology of the distant future, and yet, it’s been manning cars on the road since 2015. Insane.

3. Theory of Mind Machines

Theory of mind is the cognitive ability that allows humans to infer and understand the thoughts, beliefs, intentions, emotions, and desires of others. This is considered the next frontier of AI as it would give it the understanding of other entities that exist within the world. With this capability, they’ll communicate more effectively and will be able to better anticipate the needs and intentions of others. Basically, AI in its current evolution of limited memory machines, only relate and compare input to their stored data to generate an answer/outcome. With theory of mind, it may also infer as to why you’re asking for that information or how you might feel about it. This means that theory of mind machines would operate on artificial general intelligence. As of today, we’ve not yet achieved this capability within AI.

4. Self-Aware Machines

In theory, the most advanced type of AI would be self-aware, possessing an understanding of the world, others, and itself. This level of AI is known as artificial super intelligence (ASI). If our AI machines achieve ASI, we’d have created sentient beings. Beings who can form their own belief systems and ideologies. ASI machines could then choose to partner with us, truly working together to usher in a new era of discovery and research. As of today, this is a distant and possibly concerning future. Concerning as ASI machines may learn to wish and feel deserving of lifestyles and belongings that don’t adhere to those that we imagined or defined for them. They might even place their value and desires above mankind’s. This unpredictable possibility is often referred to as the singularity; the point or event in which artificial intelligence has evolved beyond human intelligence and control.

Well, What Can It Do Then?

Now that we’ve defined the four types of AI we can begin to understand the roles they’ll soon be taking in our society. From brain chips that will allow users to Google search using their thoughts to even the assessment of a person’s vulnerability to fall risks, AI is and will be addressing issues and uses we never could have imagined by 2024. In our next blog we’ll explore a few of the incredible (and unexpected) ways that AI is being researched, developed, and even used for our betterment as a civilization. We’ll learn how AI is being used to improve medicine, meteorology, exploration, and more.

– Daniel