There are some ‘technical members’ of the AI family which I want to very quickly mention. Reactive Machines have a self-explanatory name, they are reactionary and not the brightest star in the sky. They blurt out an immediate response to prompted questions and they have no ability to create or store data. An example would be chatbots on websites. (I personally don’t believe these bots deserve any title that includes reference to AI.)

Limited Memory AI can store past data and use that data to make basic decisions or predictions. It has a short-term knowledge base. An example would be self-driving cars.

Theory of Mind AI refers to a type of AI that can perceive our emotions and this type of AI does not exist yet.

AI of Reinforcement Learning is a system that learns through trial and error, which involves a system of rewarding desired behaviours or being punished by negative results. Imagine the AI getting the digital equivalent of a treat or a slap on the hand. I know this might sound like an inhumane method to teach, but it is similar to how young children are taught and in many ways hardwired into our brain.

The last one I want to reference is Reinforcement Learning AI, which distinguishes itself from other machine learning methods like supervised and unsupervised learning. It focuses on learning optimal actions through exploration and exploitation. (Doesn’t this remind you a little of Darwin’s survival of the fittest?)

Inputs. Humans have ‘input devices’: eyes, ears, touch, taste and smell. Our brain receives the signals from these gadgets, interprets and then decides how to react. Our noggins ‘wiring’ is called a neural network which develops in the womb using our DNA, but once we are born, our brain creates new ‘wires’ based on environmental factors. We are programmed to reprogram ourselves.

AI needs input devices. One of the most interesting developments for me at the moment is something called Computer Vision AI. ‘Narrow AI’ is using this already, and it is trained to understand video and images. We will want to give AI as many ‘senses’ as possible. Unlike us, AI can have infinite input devices.

The devices on the Internet of Things exceeded 17 billion at the time I wrote this AI Manual. Imagine being able to hear, smell, touch and see all over the world all at the same time? That is a lot of data!

You might want to know some of the ‘types of data’ nerds’ talk about? (Just encase you get stuck in a conversation with them). Training data is what helps the AI models learn. Validation data helps ‘tune’ the AI. Test data helps assess the AI model’s performance.

A hefty side nugget: Our brains are the most complicated ‘object’ we know of in the universe. It has 89 billion neurons, each connected to around 7000 other neurons that send between 10 and 100 signals every second.

The most powerful computer in the world is called the Frontier which you can find at Oak Ridge National Laboratory in the USA. This    supercomputer has a total of 8 699 904 combined GPU and CPU cores.

Our brains ‘processing data power’ has been estimated at 1 exaflops which is the same as the Frontier. The cool thing is that the Frontier uses millions of times more energy to process information than our own brains, this means ‘our computer is more efficient. Our brain also doesn’t take up an entire floor in a building. “Well done mother nature, good job!”.

If an ASI does become aware of our brilliant hardware, it might just decide to harvest a few thousand of our brains to boost its own ability! This might sound far-fetched, but technology of the future will include organic parts.)

I am confident by now you have a grasp on what AI is. Let’s talk for a brief moment about AI and you.