Virtual Intelligence versus Artificial Intelligence

 *This article was published in contribution to the new frontier of Intelligent Reality (IR). Click here to learn more about the 2022 IEEE 2nd International Conference on Intelligent Reality (ICIR).


Virtual Intelligence versus Artificial Intelligence

Like any advancing area of technology, the field of Artificial Intelligence is full of a dazzling array of technical terms, with new terms being created almost daily. Some are simple to understand, such as Narrow AI (also called Weak AI), and Strong AI, sometimes called General AI. But many require far more explanation, from the definition of a neural network to terms like computer vision and natural language processing, deep learning and augmented intelligence.

Let’s start with the basics. What do we mean when we say “Artificial Intelligence”? Probably the simplest definition of artificial intelligence is simply intelligence exhibited by a machine. While the definition of intelligence is also tricky, in this context it is generally accepted to mean that the machine is capable of taking actions based on data without explicit programming to take that action. In a traditional computer program, a robot might be programmed to walk towards a goal if given a command to walk. The walk command would include the direction, details of how to take each step, timing and many other details. In an AI system, the robot might “understand” how to walk, and be instructed to reach the same goal. However, the AI system would not be given explicit instructions on the direction to travel in, how many steps to take, or many of the other details a non-AI system would need. The AI will fill in those gaps based on the available data.

AI technology is increasingly used in many areas, not just robotics. Millions of people interact with artificial intelligence on a daily basis through digital assistants like Siri and Alexa, or recommendation systems as used by Netflix. Artificial Intelligence helps plan our routes in Google Maps, translate languages, corrects what we write, decides what social media to show us, and tells self-driving and self-parking cars how to operate. One area that has incorporated artificial intelligence for decades is video games. From pathfinding to creating autonomous virtual opponents, video games have used forms of AI since the 1950s!

Unsurprisingly, Artificial Intelligence also has many uses in virtual reality. And that brings us to the topic of this article. A rising term in the world of artificial intelligence is Virtual Intelligence. But what is Virtual Intelligence and how does it differ from Artificial Intelligence? This article will explore both. Read on to find out more!

Types of virtual intelligence and artificial intelligence

A good question to ask is what are the different types of artificial intelligence. Generally, AI practitioners agree that there are four basic types: reactive machines, limited memory machines, theory of mind machines and self-aware machines.

Reactive machines are the most basic type. They can react to different stimuli, but they do not “learn” from their experiences. They have no memory, so they don’t improve over time. They simply apply the same rules again and again.

Limited memory machines add the ability to learn from experience. Nearly all current AI systems, such as natural language processing and deep learning, are trained using large volumes of data. The AI system stores this data and uses it to complete the task it was developed for. An AI designed for image recognition might be trained on millions of images so that it can recognize any new object. The more training data the system has, the better its results will be. However, it is critically important that the AI is trained with good data, not data that reinforces biases and stereotypes. The responsible training and use of AI is a topic of active research and debate within IEEE and many other groups.
Theory of Mind machines are the next step in AI research. As far as we know, no-one has reached this level of AI yet. A Theory of Mind machine will be able to better understand the entities it is interacting with by discerning their needs, emotions, and thought processes.

The fourth type of artificial intelligence is the fully Self-Aware machine; a machine that can truly think for itself.

Some in the industry have proposed three other categories of artificial intelligence: Artificial Narrow Intelligence (ANI), Artificial General Intelligence (AGI), and Artificial Superintelligence (ASI). ANI systems refers to artificial intelligence that can only perform a specific task. All current implementations of AI fall into this category. An ANI may be great at language translation, but it would have no clue how to drive a car. AGI refers to a general purpose AI, a system that could autonomously build competencies across multiple areas. Artificial Superintelligence is an AGI that far outperforms human intelligence in all areas.

But where does Virtual Intelligence fit within this mix? What are the characteristics of virtual intelligence, and how do you distinguish between artificial intelligence and virtual intelligence?

Virtual Intelligence is an application of artificial intelligence. It simply means an artificial intelligence that is deployed inside a virtual or augmented world. Virtual Intelligence can be used in Virtual Reality, Augmented Reality, Mixed Reality or any form of Extended Reality.

In some ways, virtual intelligence follows the path of artificial intelligence in video games. An increasingly common use of virtual intelligence is to provide persistent avatars that provide information or training, and help facilitate social interactions. These avatars are effectively chatbots, but realized in a virtual environment. A virtual museum might have numerous avatars that use virtual intelligence to provide insights to museum visitors, filling the same role as a tour guide or museum curator.

Interestingly, virtual environments can also be used to train the artificial intelligence itself.

Systems used in virtual intelligence and artificial intelligence

Narrow AI in the form of Machine Learning and Artificial Intelligence is found everywhere in today’s modern world. And their use is expanding. In the early days of artificial intelligence, the primary focus was the creation of expert systems. These were used in areas such as diagnostics, process control, systems monitoring, and scheduling and planning. Early expert systems were rules based, where bodies of knowledge were represented by a collection of IF THEN statements. While this may look simple, the application of large numbers of IF THEN statements to a data set can result in a high degree of complexity.

More recently, neural networks have taken the lead. Neural networks are loosely modeled on the behavior of networks of neurons in the brain. Some newer types of neural network include deep learning, supervised learning and unsupervised learning. All of these forms of artificial intelligence require big data in the form of training sets. For example, in natural language processing, the goal is to understand written or spoken language. The AI solution will be trained on a huge dataset of language. If that dataset is extensively labeled with additional information, the AI is using supervised learning. If the dataset has no labels, the AI solution is using unsupervised learning. Different forms of neural networks are used to power conversational AI and AI assistants, especially to handle the natural language processing that underpins the virtual assistant experience. Unsupervised learning in particular can be very powerful when applied to data science and big data.

A combination of ever cheaper processing power and data storage, coupled with new techniques for artificial intelligence and machine learning has made it possible to embed an AI model in all kinds of applications, from the cloud to the edge, to low end IoT devices.

Now that virtual reality headsets have become affordable, and adoption is growing, various groups have begun to combine artificial intelligence with virtual reality in the form of virtual intelligence. Most commonly, artificial intelligence is used within virtual reality to create a virtual character that can interact with anyone entering the virtual reality experience. These virtual characters, or avatars, can be quite sophisticated. They can combine the abilities of a capable virtual assistant with all of the abilities of a virtual character, providing support, guidance and interaction to people in the virtual world. Artificial intelligence has other uses in virtual reality too.

The goals of artificial intelligence and virtual intelligence are closely related. Artificial intelligence aims to recreate human-like intelligence, but its everyday goals are usually much less lofty. Most systems that implement AI are simply trying to become more usable. The same is clearly true for systems deploying virtual intelligence: they aim to make the human experience better.

Applications of virtual intelligence and artificial intelligence

Throughout this article we’ve discussed a range of applications for artificial intelligence and machine learning. We’ve also touched on how artificial intelligence can be applied to virtual reality in the form of virtual intelligence. But there are other opportunities and use cases.

In a virtual environment, natural language processing and translation can be used to enable real-time conversation between users who speak different languages. AI-enhanced design tools can help users create better representations of themselves, and also to create ever bigger and more complex virtual worlds to explore. Artificial intelligence can even be used to improve human-computer interactions too. It’s entirely possible that new developments in AI research, such as neuromorphic computing, will open up new applications for virtual intelligence.

It’s clear that virtual reality and artificial intelligence can work in tandem to deliver many benefits.

Future of virtual intelligence and artificial intelligence

Artificial Intelligence and Machine Learning have received a lot of attention from researchers and product developers, enormous amounts of funding from the venture capital community, and almost endless coverage in the media. Artificial intelligence has been hyped, doomsayers have warned of the dangers of General AI, and we’ve all been told repeatedly that AI will take away our jobs. The truth, of course, is more nuanced. Much of the general hype around artificial intelligence and machine learning is justified. AI techniques like speech recognition (especially natural language recognition), and image processing are capable of astonishing feats. While we have probably all encountered an AI chatbot that delivered a terrible experience, there are many that are consistently good and serve a useful purpose.

However, artificial intelligence faces some very real challenges. Some, such as the enormous power requirements, are technical in nature. But many are not. It has become increasingly clear that bias in training data creates biased artificial intelligence. In a world where adding “artificial intelligence” to any idea can improve the odds of receiving funding, there continues to be an alarming number of pseudo-science applications for AI. Examples include systems that predict criminal intent or personality traits from an image of a person. These are little more than modern-day phrenology.

Another challenging area for artificial intelligence is its societal impact. While AI will not eliminate all jobs, it is likely to have a similar or larger impact than other advanced automation techniques. Unlike previous waves of technology, artificial intelligence could impact both the bottom end of the labor market and the top end. Routine tasks like sorting, packaging, and answering customer service calls could well become fully automated with AI. But at the high end of the job market, advanced translation and the drafting of legal agreements could also become largely an occupation for artificial intelligence.

The future of artificial intelligence seems almost unlimited. The endless list of possible applications—coupled with ever improving AI techniques and many new types of AI algorithm— is driving adoption and use. Cloud computing continues to increase in power and capabilities while decreasing in cost. Big data is everywhere. Artificial intelligence’s abilities in predictive analytics, AI chatbot services, image processing and natural language processing are developing rapidly. Another exciting area is generative AI for design.

While there are some very real limitations to current artificial intelligence models, and no clear path to Artificial General Intelligence or human intelligence, AI technology is increasingly available, affordable and widely deployed. While virtual intelligence is in its infancy, it’s potential is enormous and increasing daily!


2022 IEEE 2nd International Conference on Intelligent Reality (ICIR)

 Want to learn more about Intelligent Reality? Why not register to attend IEEE's International Conference on Intelligent Reality (ICIR). The IEEE International Conference on Intelligent Reality aims at identifying the challenges and opportunities inherent in deploying intelligent tools and interactive disruptive technologies into immersive environments. It provides a forum for leading researchers, industry professionals, and standards experts to share their research findings and ideas.