Artificial intelligence a.k.a. AI is no longer a concept constrained by fiction. It is alive! Humankind has already started exploring AI and is making significant improvements with each passing day.
Nonetheless, self-aware, corrupt AI-powered entities that have extensive knowledge and are hell-bent on wiping out humanity all to save humanity from itself - as we’ve seen several times in sci-fi novel and movie plots - is even much farther than a remote possibility.
But yes, AI is here, and it is doing pretty well to help mankind achieve things that previously appeared unachievable or impractical. Today, we have several practical instances of artificial intelligence, such as:
- Autopilot feature in commercial flights
- Chatbots
- Intelligent virtual assistants, like Alexa, Google Assistant, and Siri
- Plagiarism checkers
- Robots
- Self-driving cars
- Spam filters
Artificial intelligence or machine intelligence refers to any form of intelligence displayed by machines. It is the field of research that deals with simulating human-like intelligence in machines. Now, AI is of many types. But before discussing that let’s compare it with machine learning first.
Caution! AI is Not Machine Learning
Although artificial intelligence and machine learning have several aspects in common, they are not the same. There are several differences between AI and machine learning, which makes them related but independent fields of research.
In fact, machine learning is a subset of artificial intelligence, just like deep learning and neural networks are. In technical terms, machine learning is a way of data analysis that automates analytical model building.
ML is a branch of AI that deals with allowing machines to learn from data, followed by identifying patterns and forming decisions with minimum human intervention. AI, on the other hand, has a much wider scope with ML being an aspect of it.
You can also learn about the future of artificial intelligence.
What is Generative AI
Generative AI produces stuff. Think of it like OpenAI's ChatGPT and GPT4. Think of it like Midjourney. These are the types of artificial intelligence that create text, images, video, or code.
Types of AI
AI research focuses on developing machines/systems that can emulate human-like behavior, especially in terms of thinking and performing tasks. Therefore, human intelligence is used as a comparison parameter to group the various types of AI.
Now, there are two ways in which all AI and AI-based systems can be grouped into. The first one is on the basis of capability, and the second one takes the functionality as the base for grouping AI.
1. Capability-Based
On the basis of their capability i.e. the ability to do tasks in comparison to human task performing ability, AI is categorized into 3 groups or 3 Types or Article Intellignce; Narrow AI, General AI, and Superintelligence.
1.1. Narrow AI a.k.a. Weak AI
The common form of artificial intelligence that is available today is narrow AI. Its scope is limited by the ability to perform dedicated, predefined tasks. Google Assistant, Siri, and even the IBM Watson supercomputer are examples of weak AI.
Narrow AI lacks consciousness and self-awareness, and is less intelligent than humans. As weak AI is trained only for doing a specific task or a set of tasks, they can’t perform anything beyond their defined capabilities. If forced to do so, the results become unpredictable and unuseful.
1.2. General AI
A machine or system can be considered general AI if it is able to perform an intellectual task with human-like efficiency. The purpose, driving the notion of general AI is to develop an artificially intelligent system that can replicate the smartness and thinking of a typical human.
Unfortunately, there is no existing example of general AI. In other words, general AI is still theoretical and not practical, not yet.
AI researchers, however, are now focusing on developing systems that can be labeled as general AI. The time and effort required to do so are, naturally, unprecedented.
1.3. Super AI/Artificial Superintelligence (ASI)
At the super AI level, artificially intelligent systems best, or transcend human intelligence. Such AI systems can perform tasks far better than a human and that too with cognitive properties. A few key characteristics of superintelligence systems are:
- Ability to think, reason, and make judgments
- Able to solve puzzles
- Can plan on its own
- Communicate effectively
- Learn from surroundings, such as people, environment, and incidents
ASI is a hypothetical concept that helps in exploring the endless possibilities offered by artificial intelligence. If superintelligence somehow becomes feasible, such a form of artificial intelligence has the power to transform the whole world as we know it.
2. Functionality-Based
Next, we will classify AI on the basis of functionality. This classification of AI takes into account the way in which AI systems behave and feel, also taking in their cognitive prowess and common (machine) sense.
There are 4 groups in which we can put any AI system based on its functionality; Reactive Machines, Limited Memory Machines, Theory of Mind, and Self-Aware AI.
2.1. Reactive Machines
Like Narrow AI, these are the most basic types of AI systems that we have today. Reactive machines cannot form memories and thus, cannot react to future conditions based on their past experiences.
Reactive machines can only perform their best while reacting to present conditions. Two good examples of reactive machines are:
- IBM Deep Blue - A chess-playing computer developed by IBM that became the first-ever computer chess-playing system, in 1997, to win a chess game against a reigning world champion, Garry Kasparov.
- Google AlphaGo - A Go-playing computer program that became the first AI system, in 2015, to beat a Go professional without handicap on a full-sized 19x19 board.
2.2. Limited Memory
Limited Memory AI systems are capable of storing a few past experiences or some data for a short while. This data is usable, only for a limited time period. The best example, until now, of limited memory AI systems, is self-driving cars. They are able to store information such as:
- Speed of close by cars
- Distance from those cars
- Driving speed limit
- Road directions
Self-driving cars are limited memory systems that use their memory storage capabilities to navigate around the road. This, however, doesn’t mean that they are free from errors. It is still a long way for self-driving cars to become a dependable option for common, everyday use.
Content generation remains a popular, functional use for artificial intelligence. Products like TextBuilder.ai allow site owners to quickly create full blog posts and templates in seconds.
2.3. Theory of Mind
The Theory of Mind AI refers to machines/systems that are capable of recognizing human beliefs, emotions, intents, values, etc. Such AI systems might also be able to reciprocate/respond to human emotions and/or expressions in a human-like way.
Although no substantial progress has been made in developing the Theory of Mind AI systems, researchers and professionals across the globe are working rigorously to develop such AI machines/systems.
2.4. Self-Awareness
Although self-aware AI doesn’t exist in reality, at least, for now, it is a hypothetical concept that explores AI-powered machines with thinking capability much higher than a typical human mind.
Self-aware AI systems, in theory, have their own consciousness, reason, and sentiments. They have an individuality just like any human being but are much more intelligent than any human in the world. You can read about this in many AI books, fiction and non-fiction.
Ways to Control AI
Once you've built a successful artificial intelligence, you'll need a way to keep it on the rails. This is especially true for large platforms like OpenAI and Google Bard. It's becoming much more important to develop and maintain methods for keeping your AI from producing untrustworthy (or even malicious) results.
AI Capability Control
Recently, we've been talking about AI Capability Control. That's just a fancy way of saying, "let's put rules in place to make sure our artificial intelligence does what we want it to do. . . and only what we want it to do."
In practical terms, this means adding guidelines for your AI product. Chatbots sometimes have directions to avoid providing negative, misleading, or illegal advice.
Modern hackers understand there are limits to these controls, and they've made the news recently by breaking systems to make popular chatbots go against their original design functions.
So how do you protect against this?
The most important way to build AI Capability Control is to think like a hacker. You'll need to understand the methods hackers use to break into systems, defy controls, and earn extra access. Only once you understand the inroads can you set up redirects and roadblocks.
Looking for more help with this? Note that we also put together a list of ethical hacking tools.
Conclusion
Artificial intelligence has continuously captured the imagination of humans. Concepts resembling AI have been long used by fiction writers, movie makers, and dreamers, even before AI was a thing.
Now, artificial intelligence is an undeniable reality. Although there is a long arduous way ahead of us, it is a fascinating path that might even help humans to understand themselves better and reassess many notions, including nature and life.
The evolution of AI will lead to the birth of new philosophies. It will be important for us humans, the ones responsible for conjuring these void-of-flesh-and-blood beings into our realm, to responsibly undertake the process of exploring the uncharted territory of machine intelligence.
Want to explore AI more? Check out these best artificial intelligence tutorials.
People are also reading:
- Best Artificial Intelligence Courses
- Top Data Science Tutorials
- Benefits of Artificial Intelligence
- Applications of AI
- Top Deep Learning Books
- Top 10 Data Science Books
- 10 Best Machine Learning Books
