Artificial intelligence (AI) is the science of making machines think like humans. It’s a rapidly developing field, and it’s coming to a new industry faster than you might think. AI has already been used to analyze data and predict user behavior to drive sales; reduce churn by reminding customers of their account status, and personalize advertising based on past purchases.
The term “artificial intelligence” is often used interchangeably with other terms like machine learning, natural language processing, or computer vision. However, artificial intelligence is actually just one branch of AI called artificial general intelligence – or AGI for short.
There are multiple types of AI, each geared towards solving a unique kind of problem. Understanding the type of AI your company needs can help you prioritize your efforts and avoid reinventing the wheel unnecessarily.
Weak Artificial General Intelligence (Weak AGI)
Weak AGI is a type of AI that has been around for a long time, and it’s typically applied to tasks like language processing, pattern recognition, and visual understanding. These AI systems can handle basic decision-making tasks, like identifying which of a set of sentences is relevant.
In fact, they can even solve small problems like this if they’re given a set of rules and enough examples. However, they aren’t equipped to tackle a large variety of challenging problems. Weak AGI is often used to build conversational bots that can hold basic human-to-bot conversations.
Weak artificial general intelligence is often referred to as weak AI because it doesn’t have the cognitive ability to take on more complex tasks. It’s one of the earliest types of AI, and it’s still used today in many applications. This type of AI works best when given a huge amount of data and a specific set of instructions to follow.
Strong Artificial General Intelligence (Strong AGI)
Strong AGI is similar to weak AGI in that it isn’t capable of human-level cognition. However, it does have the cognitive ability to carry out a variety of tasks that people rely on in their daily lives, like recognizing objects, understanding language, and understanding causal relationships between events.
For example, strong AGI can understand what restaurant an image represents, able to recognize a specific item on the menu. It can also understand the relationship between that item, like a hamburger or salad, and the price, like $10 or $15.
If strong AI is like a human with a Ph.D. in computer science and statistics, then narrow artificial rationality (NAR) is like a high school dropout. NAR systems have basic cognitive abilities and can be used to perform a variety of tasks. Narrow AI is often used in things like web scraping, where it’s tasked with finding data on the Internet and fetching it for analysis.
Weak Narrow Artificial Rationality (Weak NAR)
Weak NAR is a type of Artificial intelligence that can handle basic logic and decision-making. Think of it like a smartphone with a basic understanding of cause and effect and an ability to parse basic sentences. For example, it might be able to understand questions like “How many kids do you have?” or “How many cats do you own?” and return a reasonable answer.
Weak NAR can be further broken down into weakly structured and weakly understood NAR. Weakly structured NAR can understand only general concepts, like “family” or “friend.” Weakly understood NAR can understand more specific concepts, like “brother-in-law” or “feline companion.”
Strong Narrow Artificial Rationality (Strong NAR)
Strong NAR is a combination of weak NAR and narrow Artificial intelligence. It’s capable of performing basic logic and reasoning, but not with a very deep understanding of the world. For example, a person with only a high school education might understand the difference between an airplane and a helicopter, but they wouldn’t know if the helicopter was used to transport people or goods.
A person with a strong narrow Artificial intelligence understanding would be able to differentiate between the two and make the correct judgment call. Like weak NAR, strong narrow AI can also be subdivided into weakly structured and weakly understood.
Weakly understood strong narrow AI understands only general concepts, like “family” or “friend.” Weakly structured strong narrow AI understands more specific concepts, like “brother-in-law” or “feline companion.”
Summing up
These types of Artificial intelligence are at the lower end of the spectrum when it comes to cognitive ability. They can handle basic tasks like understanding language, recognizing objects, and performing basic logic.
However, they aren’t capable of solving complex problems or carrying out complex tasks like driving a car or operating a factory machine. In order to move further toward the goal of creating human-level Artificial intelligence, researchers will need to continue to explore different types of Artificial intelligence and develop better methods of analyzing data.
