Artificial intelligence (AI) has been on everyone’s lips for some years now. It’s a major technological advance, transforming the way we work and revolutionizing our daily lives. However, did you know that this constantly evolving technology can be classified into four different types, depending on its functionalities? Indeed, not all artificial intelligences have the same capabilities. While some, like reactive AI, only react to immediate stimuli, others, like theory-of-mind AI, could manage to understand and interpret human emotions. Let’s find out more about the characteristics and functioning of these four types of AI.
1. Reactive AI
Reactive AI is the first type of artificial intelligence, and the one with the most basic characteristics. Its operation is based on real-time data analysis. AI is capable of responding to immediate stimuli, but it has no memory and is therefore unable to learn from past experience. Although this is a limited form of artificial intelligence, it remains highly effective in a variety of circumstances.
Let’s take a closer look at how reactive AI works:
- Reactive AI reacts by observing its environment and according to a predefined set of rules.
- This AI doesn’t store any data, nor can it anticipate future events.
- It is designed to perform specific tasks, and is unable to adapt to new demands.
This is a form of AI frequently found in our daily lives, notably in the form of basic voice assistants, developed to respond to specific commands, without understanding the context or intent of the requests. We also find it in basic facial recognition systems, capable of identifying faces by comparison from a fixed database. Another example: IBM’s Deep Blue supercomputer, which beat chess champion Garry Kasparov in 1997, is a machine that uses reactive AI.
What are its limits?
Despite its effectiveness in solving specific tasks, reactive AI has its limitations. First of all, it is not a learning AI, which means it offers no possibility of evolution. It cannot perform tasks that require memory or in-depth understanding of context, and its use is restricted to predictable environments, where every action is linked to a limited set of possible responses.
2. AI with limited memory
Memory-limited AI represents a major advance on reactive AI. This is the most modern type of AI known today, an artificial intelligence that learns from experience in order to improve its performance, with the ability to store a variety of information temporarily.
How does this AI work?
- AI with limited memory collects data that enables it to adjust its behavior accordingly.
- It is able to draw conclusions based on past experience.
- Its learning is generally based on advanced artificial intelligence models such as machine learning and deep learning.
Although this type of AI is less common today than its reactive version, it is gradually gaining ground in our daily lives. It is, for example, the type of AI used by the most advanced chatbots, capable of memorizing our past interactions so as to improve the responses offered. Autonomous cars are another good example, taking advantage of sensors and analysis of past traffic data to adjust their driving in real time, and lead humans to their destination.
What are its limits?
The main advantage of AI with limited memory is its ability to improve its performance at any time. Its stored data enables it to adapt to new situations rapidly, a particularly interesting feature in certain fields that require real-time analysis, such as healthcare, finance or transport.
It also has certain limitations. First of all, its memory is limited, and can only store certain types of information for a limited period of time. To remain functional, it needs to be updated regularly, and it has no emotional intelligence.
3. AI based on theory of mind
This type of AI is the subject of a great deal of research, although to date no one has succeeded in developing an AI with emotional intelligence. This theory-of-mind-based AI should be able to understand and interpret human emotions, intentions and beliefs – not the easiest thing in the world to put into action!
The keys to how Theory of Mind-based AI works:
- Beyond simply processing the data it receives, this AI attempts to understand human behavior.
- It is designed to develop more natural interactions with humans, through the recognition of tone, voice, facial expressions and the general body language of the interlocutor.
- For this to happen, major advances in neuroscience and cognitive psychology are required.
This type of AI could prove useful in many circumstances, notably in the field of customer service, by proposing interactions adjusted to customer reactions, be they expressions of frustration or satisfaction. It could also enable the development of care robots capable of responding empathetically, by interpreting the emotional state of patients.
What are its limits?
The development of such AI remains difficult, which is its main limitation today. How can we accurately model human emotions? AI based on the theory of mind can also raise ethical questions, as it still needs to be accepted by everyone and committed to respecting certain limits in terms of privacy.
4. Self-aware AI
Finally,self-aware AI is the most advanced AI model imagined to date. Its performance takes it to the next level, with true consciousness, self-understanding and the ability to reflect on its own existence. In short, this artificial intelligence would be capable of reasoning in the same way as human beings. Rest assured, this is pure science fiction at the moment!
What would its characteristics be?
- It’s an AI capable of making autonomous decisions, based on its own thinking.
- It would have its own personality, its own motivations and its own will.
- These abilities would enable it to develop more authentic interactions with humans.
This type of AI does not yet exist, but many examples can be found in science fiction. Examples include the AI HAL 9000 from the film 2001: A Space Odyssey, and the artificial intelligence Skynet from the film Terminator.
What are its limits?
Although this AI presents promising capabilities, including the ability to perform tasks completely autonomously, a broad understanding of human needs, or its ability to solve complex problems, it could also face various challenges. As in the worst science-fiction scenarios, what would happen if this AI developed goals contrary to human interests? This form of AI would need to be properly framed, not least from an ethical and philosophical point of view.
To conclude
Artificial intelligence is an exciting and rapidly evolving field. Each new advance facilitates the design of more sophisticated intelligence, capable of responding even more precisely to human needs and expectations. Although today, the only fully exploited AI models are reactive AI and AI with limited memory, research is continuing into the design of AI based on theory of mind, with a view to perhaps one day succeeding in creating the first self-aware AI.
