What Are the 4 Types of Artificial Intelligence?
There are four recognized types of artificial intelligence: reactive AI, limited memory AI, theory of mind AI, and artificial general intelligence (AGI). The first two exist and are deployed today. The last two remain theoretical. Here is what each type means concretely for a business leader.
Type 1: Reactive AI
This is the most basic type. Reactive AI stores nothing. It receives an input, it produces an output. No context, no history, no learning.
The most well-known example: Deep Blue, IBM’s program that defeated Garry Kasparov at chess in 1997. It analyzed the position on the board and calculated the best move. It remembered nothing from previous games.
In your business, this is the basic recommendation engine that displays “customers who bought X also bought Y.” Effective. Limited. Predictable.
Type 2: Limited Memory AI
This is where we are today. The vast majority of AI systems deployed in businesses belong to this category.
These systems learn from historical data. They improve over time. They retain recent context to refine their responses.
Concrete examples:
- GPT-4 and its successors: they use the context of your conversation to respond coherently.
- Bank fraud detection systems: they compare each transaction against thousands of past transactions to detect anomalies.
- CV screening tools I observe with my recruitment clients: they learn from retained profiles to refine future selections.
Maroc Cloud recently launched Gemini Enterprise in Morocco. This is the type of AI, limited memory, that Moroccan companies are beginning to deploy today.
If you want to understand how to choose the right tool for your organization, I analyzed the available options in my article on the best AI for businesses.
This type of AI also raises the most urgent governance questions. According to Le Matin.ma, the European AI Act is beginning to catch up with Moroccan players who work with European partners or markets. If your company falls into that category, these rules apply to you regardless of where you are based.
I have built a diagnostic framework to assess your readiness for these requirements. Download the Board Pack AI 2026.
Type 3: Theory of Mind AI
This type does not yet exist. Not commercially. Not in laboratories at an operational level.
The concept: an AI capable of understanding the emotions, intentions, and beliefs of a human interlocutor. An AI that adapts its behavior not to data, but to a perceived mental state.
Some researchers believe that models like GPT-4 show early signs of this capability under very specific conditions. But this is not theory of mind in the strict sense. It is highly sophisticated statistical simulation.
For a business leader, the practical question is simple: do not make purchasing decisions based on a promise of “emotional understanding” from an AI tool. That is not where we are.
Type 4: General AI and Superintelligence
Artificial general intelligence, or AGI, refers to an AI capable of performing any cognitive task a human can accomplish. Superintelligence goes further: an AI that surpasses humans in every domain.
Sam Altman of OpenAI speaks of AGI as a near-term horizon. Other serious researchers estimate we are decades away. The debate is open.
What is certain: no company should build its strategy on the assumption of imminent AGI. Build on what exists. Anticipate what is coming.
What is coming is an acceleration of type 2 capabilities. Models are becoming more powerful, more specialized, more integrated into business processes. That is where real competition between companies is playing out today.
As I explained in my analysis of AI in recruitment in 2026, organizations that gain an advantage do not do so through magical technology. They do it because they structured their approach before others did.
What This Means for You
You do not need to master academic taxonomy. You need three reflexes.
First reflex: when a vendor presents you with an AI solution, ask them which category it falls into. If the answer is vague, be cautious.
Second reflex: everything deployable today belongs to type 2. Evaluate it on concrete criteria: training data quality, decision traceability, regulatory compliance.
Third reflex: types 3 and 4 are strategic thinking horizons, not operational realities. Do not put them in your 2026 budget.
If you want to go further on building your team’s AI capabilities, read my guide on AI training for business leaders.
If you are a CEO or CHRO and want to structure your AI approach on solid foundations, request a free diagnostic.
FAQ
What is the difference between weak AI and strong AI?
Weak AI refers to systems designed for a specific task, such as recognizing an image or translating text. This is everything that exists today. Strong AI, or general AI, refers to a system capable of reasoning about any problem like a human. It does not yet exist.
What type of AI is ChatGPT?
ChatGPT is a limited memory AI. It learns from large amounts of text, retains the context of a conversation, and generates coherent responses. It has no consciousness, no emotions, and does not understand in the human sense of the word.
Is artificial general intelligence coming soon?
Some OpenAI executives mention a horizon of a few years. Independent researchers speak of decades. For a business leader, the right posture is to build on current capabilities while staying informed about developments.
Which types of AI are regulated by the European AI Act?
The AI Act applies primarily to limited memory systems deployed in high-risk contexts: recruitment, credit, healthcare, security. If your company uses these systems and operates with European partners, you are within the regulatory perimeter, even if you are based in Morocco.