You've probably heard how AI is revolutionizing the way marketers work. In fact, you may be using AI tools right now.
But if you, like me, have not yet "pulled the curtain" to see how this technology works - then it is necessary to read the article.
Here, we'll cover the four main types of AI – reaction machines, finite memory, theory of mind, and self-awareness – and how each type powers your marketing.
How many types of AI are there?
There are four main types of artificial intelligence: reaction machines, finite memory, thinking theory, and self-awareness.
However, since AI can be categorized by function (the types listed above) and capabilities, you can add three more: narrow intelligence (ANI), general intelligence (AGI), and superintelligence (SGI).
Below we will explain each type.
Reaction machine
As the name suggests, reactive machines react to different cues. It does this without using memory or a broader understanding of context.
For example, this type of AI is often used in game design to create opponents. Opponents react to your actions or attacks in real time, but don't know the overall goal of the game. In addition to that, it doesn't store any memories so it doesn't learn from past experiences and tweak the gameplay.
Reactive AI powers many marketing tools. Chatbots are a notable example. These programs use reactive artificial intelligence to respond to messages (or inputs) with the right information.
Chatbots are a popular tool in customer service, but they can also increase the productivity of marketers. HubSpot's ChatSpot, for example, is a handy AI assistant that can pull reports, create contacts, and send follow-up emails based on certain commands.
In addition to chatbots, reactive AI can analyze customer behavior, campaign performance, and market trends. With these insights, marketers can optimize their campaigns at any time, increasing their efficiency and ROI.
Limited memory
Limited in-memory AI is able to learn from a limited amount of data or feedback. However, it does not "store" any memories for a long period of time.
A good example of this "limited" aspect of AI is ChatGPT. It has a limit of 4000 markers (in the form of text like words) and cannot recall anything from the current conversation after this limit. So, if the session is 4097 tokens, ChatGPT responds based on the latest 97 tokens.
This technology can be applied in self-driving cars. It can detect lanes and map the road ahead. It can also adjust vehicle speed and brakes in real time based on traffic patterns and road conditions.
In marketing, AI with limited memory can be used to analyze large amounts of data to help marketers make more informed decisions about their strategies and tactics. It can also make predictions and recommendations based on this data.
While finite memory algorithms are effective, they are not foolproof. They can make mistakes or provide inaccurate predictions, especially when dealing with outdated data. In other words, the output works as well as your input. Therefore, it is important to train these algorithms with accurate, relevant, and up-to-date information.
Reactive machines and limited memory AI are the most common types today. They are both a form of narrow intelligence (which we will discuss further below) because it cannot go beyond programming capabilities.
Theory of Mind
Theory of mind exists only as a concept. It represents an advanced technology capable of understanding the human state of mind.
For example, if you yell at Google Maps because you missed a turn, it doesn't feel offended and doesn't offer emotional support. Instead, it responds by looking for another route.
The idea behind the theory of mind is to create machines that can interact with humans more effectively because humans understand their own needs, goals, and motivations. For example, if an AI system can understand the frustration of disgruntled customers, it can react more subtly.
In the long run, the theory of mental AI could have a significant impact on marketing. However, it is still in its early stages, and it is difficult to predict when it will become a reality.
self-awareness
Self-aware AI is seen as the next stage in the evolution of theory of mind, with machines able to understand human emotions and have their own emotions, needs, and beliefs. Currently, this type of AI only exists in hypotheses.
The robot M3gan in the movie of the same name is an example of self-aware artificial intelligence. She is sentient, knows who she is, experiences emotions and is able to understand the emotions of those around her. She will be embarrassed and will also have social interactions.
The stage of artificial intelligence
AI has three phases, largely defined by its ability to replicate human capabilities:
1. Narrow Artificial Intelligence (ANI): Narrow AI represents most AI systems that exist today. At this stage, AI is designed to perform a specific task or set of tasks. It has no ability to learn or adapt other than programming. Examples include chatbots and virtual assistants such as Siri, as well as recommendation algorithms.
2. General Intelligence (AGI): This is the next evolution of artificial intelligence. These systems are designed to have human-like intelligence that allows them to learn and adapt to new situations, think abstractly, reason, and solve problems. For now, AGI is still largely theoretical.
3. Superintelligence (ASI): ASI is a form of advanced artificial intelligence that goes beyond human intelligence, enabling it to solve complex problems, create new technologies, and make decisions beyond human comprehension. ASI is a hot topic, and its potential benefits and risks are highly speculative.
While these stages are widely accepted, there is ongoing debate about the definition of each and when we can achieve them, or whether we should evolve AI.
Top AI in marketing
As mentioned earlier, both reactive AI and finite memory AI (both narrowly defined) exist today. This means that the AI tools marketers use are strictly passive, or passive + limited memory.
We surveyed more than 1,350 marketers in the U.S. to understand their use of AI and automation, as well as the tools they use in their roles. Here are some key points.
First, when asked about generative AI tools used in marketing personas, most marketers use AI chatbots (66%).
Chatbots can be both reactive and limited memory AI. For example, a rule-based chatbot that follows an if/then model and is programmed with a fixed response can be called reactive AI because it follows a set of structures and cannot deviate from them.
Machine learning chatbots, like conversational chatbots, are memory-limited AI because they use data and past conversations to respond to customers. Over time, they become more effective, but they have a limited memory.
Marketers also say they typically use visual AI tools (57%) and text generation tools (56%). Regardless of the tools they use, all generative AI is finite memory AI because these tools can create new content based on the trained data.
All AI/Automation users who responded to our survey reported that AI and automation tools saved an average of 2 hours and 24 minutes per day.
From reactive machines to finite memory AI, theory of mind, and self-awareness, each type of AI has its strengths and limitations. Understanding these differences is key to choosing the right tools, utilizing them effectively, and staying ahead of the curve.