Request a free audit
Take advantage of our complimentary, no-obligation complete audits.
Social media audit Problem definition workshop ISO/SOC readiness
7 types of Artificial Intelligence (AI)
Explore the fascinating world of AI, from Narrow to Super AI, and understand how these technologies are shaping our future.
Key takeaways
- AI Complexity: AI encompasses a range from simple data analysis to understanding human emotions, divided into capabilities and functionalities.
- Capabilities: AI is categorized into Narrow AI (task-specific), General AI (human-like abilities), and Super AI (surpassing human intelligence).
- Functionalities: Includes Reactive Machines (basic responses), Limited Memory (data-driven improvements), Theory of Mind (understanding emotions), and Self-awareness (conscious machines).
- Future Outlook: While the potential for AI to evolve into Super AI exists, its current and near-future applications are focused on augmenting human capabilities and solving complex problems.
Classifying AI
Most people have a basic understanding of Artificial Intelligence (AI). However, an all-encompassing definition of AI is still a challenge due to the complexity of all its functions and purposes.
AI systems range from gathering data and analysing basic patterns to understanding complex human emotions. Different types of AI – with different subcategories and functionalities – are used to achieve a broad range of outcomes.
To make it simple, artificial intelligence can be divided into two broad categories: AI based on its capabilities and AI based on its functionalities.
AI-based on capabilities
Simply put, AI is a form of machine intelligence designed to mimic human decision-making and problem-solving capabilities. All AI types use machine learning, deep learning, and neural networks to complete specific tasks. The degree to which they utilise these technologies inform their capabilities and thus can be placed into 3 levels.
Narrow AI
Artificial Narrow Intelligence (ANI), also known as Weak AI, is task-specific and designed to perform only one basic function. This fit-for-purpose technology performs better than humans in narrowly defined, structured tasks with limited parameters and contexts, such as internet searches (Google’s RankBrain) and voice recognition. We are currently living in the age of Narrow AI.
Examples of Narrow AI are the image recognition systems of self-driving cars, smartphone assistant speech and facial recognition, and Netflix’s recommendation lists based on user activity and preferences.
General AI
Artificial General Intelligence (AGI), referred to also as Strong AI or Deep AI, is where machines have the ability to think, learn, understand, and make decisions like humans. General AI allows machines to apply knowledge and skills to various contexts to solve complex problems.
AGI uses a theory of mind AI framework (we’ll touch on this later) to learn, understand, and recognise the emotions, beliefs, and thought processes of other intelligent systems. Although General AI is still nascent, it is where we are headed. Famous fictional examples of what General AI could look like are WALL-E, the humanoids in Blade Runner, and HAL 9000 from the film 2001: A Space Odyssey.
Super AI
Artificial superintelligence (ASI), also known as ‘help!-the-machines-are-taking-over!’, is the final form of AI. Super AI will outperform humans in every aspect, surpassing our intelligence to perform any task harder, better, stronger, and faster. This is also the stage at which the AI Control Problem becomes key for our society since we want ASI to be aligned with our goals and values.
ASI systems will be able to comprehend complex and nuanced human emotions, experiences and sentiments. Not only this, but it will also be able to generate emotions, beliefs, and desires of its own. Yikes.
But don’t worry – yet. Super AI is still only hypothetical. The era of machines having superior problem-solving and decision-making capabilities to make judgments and independent decisions is far off. Yet, some tech entrepreneurs predict we’ll likely reach the singularity in 7-30 years. Double yikes.
AI-based on functionalities
Now that we understand the capabilities of the various types of AI, let’s take a look at its different functionalities.
Reactive Machine
Reactive machines are the most basic type of AI. As the name implies, these machines are reactive to stimuli and respond to immediate tasks and requests with predictable outcomes. They only function when given limited inputs and are incapable of storing memory or improving their functionality through learning or experience.
Their ability to read and respond in real-time allows them to perform basic autonomous functions. This includes email spam filtering, YouTube recommendations, and even beating a chess Grand Master.
Limited Memory
Building upon Reactive Machines, Limited Memory was created to excel in more advanced scenarios thanks to developments in memory management and data storage. Limited memory is able to store past data and make predictions from it by gathering learning data and improving over time.
Almost all current applications we use today utilise Limited Memory AI. They are all trained by large datasets that are stored to create a reference model for solving future problems. Chatbots, automated vehicles, and facial and voice recognition are all powered by Limited Memory AI.
Theory of Mind
Theory of Mind AI is where machines are able to truly understand every human aspect, from picking up on subtle environment changes and reading emotional cues. It means that machines will be able to generate their own meaning and understanding rather than simply replicating the human mind.
The idea of “theory of mind” comes from psychology which describes a human’s ability to read the emotional cues of others and adjust behaviour to predict future outcomes based on that information.
Although this type of AI has not yet been fully realised, we are pretty close. Kismet, a robot created in 1997, showed a few aspects of Theory of Mind AI. It was able to recognise human emotions and could mimic them using its mechanical facial features. Additionally, the humanoid robot Sophia (2016) is also able to see our emotions and respond appropriately.
Self-awareness
The final stage of AI development will be its ability to become self-aware, much like (most) humans are. Self-aware AI will create machines that deeply understand and are aware of complex human emotions and mental states, including their own. This type of AI will have consciousness and emotional quotient that equals – and perhaps overtakes – that of human intelligence. Self-aware AI will experience the same wants, needs, desires and emotions as humans.
But Harrison Ford and Ryan Gosling won’t be hunting down replicants in a Blade Runner dystopia just yet. We simply don’t have the technology currently. But Artificial Superintelligence will surely power the hardware and algorithms when they are developed.
Summary
Will machines take over the world as our favourite sci-fi authors would have us believe? Or is the destiny of AI firmly rooted in helping and collaborating with humans to keep society advancing?
It is not entirely implausible to consider there might be more types of AI that will soon be developed in partnership with advanced AI. As we continue to push the capabilities of this powerful technology, only time will tell if we are destined for singularity – or whatever comes after.
Mario Grunitz
Working Machines
An executive’s guide to AI and Intelligent Automation. Working Machines takes a look at how the renewed vigour for the development of Artificial Intelligence and Intelligent Automation technology has begun to change how businesses operate.
