What is AI? Artificial intelligence was the most searched technology of 2023, even earning Word of The Year in the Collins Dictionary. With hundreds of millions of users interacting with AI technologies such as chatbots every week, it’s important to define what artificial intelligence is, and what it isn’t. We’ll also look at what AI can do, and how it’s being used today.
What is AI? Artificial intelligence explained
Artificial Intelligence is a subset of Machine Learning (ML), which itself is a subset of Computer Science (CS). AI is the simulation of intelligent behavior, using computers.
One of the most popular subsets of AI is natural language processing (NLP), the simulation of language-based communication, using computers. ChatGPT is an example of artificial intelligence that uses NLP, because it communicates with the user by using the same language with which a human would communicate with another human via a computer.
The four types of AI
AI can be categorized by degree of scope and power, with terms such as weak, strong, AGI, and ASI. The below categories are approximate to the reactive, limited memory, theory of mind, and self-aware classifications of AI.
Weak AI or Narrow AI
Weak, or narrow, AI is AI designed for a specific purpose. It can perform specific tasks, but not learn new ones. Language translators, virtual assistants, self-driving cars, AI-powered web searches, and spam filters are examples of weak or narrow AI. It is formally known as artificial narrow intelligence (ANI).
Some but not all weak AI systems involve deep learning algorithms. If a deep-learning algorithm is involved, it will self-improve over time to become better (faster and/or more accurate) at the task than its human creator. The alternative to a deep-learning algorithm is a machine-learning algorithm with only one layer of parameters. In this case, it will be trained to be proficient at a task, and then remain at that level of proficiency. This could still be better than a human in terms of speed and accuracy but is not what everyone’s excited about; Deep learning is the fun part.
Strong AI or Broad AI
Strong AI, formally known as generalized AI, can perform many tasks. This potentially includes tasks that were unforeseen by its creator. It can also learn new tasks. ChatGPT is now an example of generalized AI. There are hundreds of plugins that each expand the functionality of the chatbot beyond what was intended by the programmers who created it. Broad AI will perform tasks using data outside of its own training data. ChatGPT is a special example in that it has access to the internet via the Bing search engine.
Using a deep-learning algorithm, it will self-improve over time to become better at the task than its human creator.
Essential AI Tools
Content Guardian – AI Content Checker – One-click, Eight Checks
Jasper AI
WordAI
Copy.ai
Writesonic
Artificial General Intelligence (AGI)
This is AI with human-level consciousness, which exhibits self-awareness and human intelligence. Sometimes called Super AI or Conscious AI, artificial general intelligence will be capable of performing as many tasks as a human (an infinite list, in theory) as well as prioritizing those tasks, and learning new ones. It will also understand why it is performing and prioritizing them, giving it the agency to make independent decisions.
Achieving AGI will be one of the most significant points in our history. This theoretical point is called the singularity, at which point we will prove that an intelligence can create an intelligence equal to itself. However, there are critical ethical problems to be solved before any can safely create AGI, known collectively as the alignment problem. This is because creating AGI will almost inevitably lead to ASI, or Artificial Super Intelligence.
Artificial Super Intelligence (ASI)
Artificial Super Intelligence is AI of above-human-level intelligence. Sometimes also referred to as Super AI or Conscious AI, it’s best to specify either AGI or ASI to avoid confusion with the other. Should humans create ASI — an intelligence more intelligent than ourselves — that AI could, in theory, create an AI more intelligent than itself.
At this point, we would not necessarily have control of the intentions and objectives of this superior intelligence. It would then be limited only by the bandwidth and processing speeds of current hardware. AI already exists for molecule discovery in material science and optimization in computer science. Considering this, it’s reasonable to expect that a Super Intelligence would independently choose its own tasks, optimize its own efficiency at those tasks over time, and also optimize the hardware that it runs on if given access to the robotics required for manufacturing.
FAQ – What is AI in simple words?
The simplest way to explain AI is that it involves using computers to complete tasks that would otherwise require a human. Artificial intelligence is, after all, the simulation of human intelligence using computers.
There is a significant difference between AI system and traditional computer systems. To exemplify this, let’s look at what non-AI systems can do, and what they can’t. Traditional computer systems can complete data-based tasks faster than a human. For example, a calculator can perform the sum 781 x 3,984,159 faster and with less error than a human brain. They can also retrieve a greater variety of information than a human. For example, Google will know the names of every city, and every town of every country in recorded history — a memory feat above that of a human. These tasks are not what AI is for.
Generative AI does something no other kind of computer system does. It doesn’t sum pre-existing numbers, and it doesn’t retrieve pre-existing information. ‘Gen AI’ creates new and unique data — new sentences, songs, images, and more — that did not exist before. By learning from examples, it can learn what a “good photo” looks like in the same way as a human photographer. Generative AI will then create a new set of pixels that fits the invisible rules we accidentally use when deciding which photos are good, and which aren’t.
