“Is AI a threat to humans?” Ludens World and The role of “human-like” in the “Virtual Human”

There have been many discussions turning up saying that by the year 2045, AI will surpass humanity. As the boundary between human and technology steadily vanishes, what will be the role of the “next generation interface”?

Towards an interface that humans can understand

The world’s first computer, ENIAC, was the size of two public elementary school classrooms (167m2), and it was incredibly difficult to figure out how to operate it in order to get your desired results. Since then there has been an increasing demand for more intuitive and efficient computers, and various human interfaces have been created to bridge the gap between humans and computers in order to fulfill that demand.

Most early computers executed commands through a batch system, in which several instructions were given through a punch card and then processed all at once; in order to improve productivity and responsiveness, the inputting and outputting of information was gradually changed into a real-time interactive system. This included the introduction of displays that you interact with by drawing pictures, as well as keyboards and the mouse, giving computers a human interface that anyone could easily learn to use.

Then, since computers based on Moore’s Law became high-performing and compact, the concept of the personal computer ー that anyone could access a computer anytime, anywhere ー was born. In addition, devices capable of 24-hour connection such as smartphones and tablets, like the iPod, and smart speakers, like Alexa and Google Home, have also been introduced. In other words, the evolution of human interfaces such as touchpads and voice inputs has also continued to progress.

The interfaces are characterized by having specialized in the ability to communicate necessary information at the necessary time, and then getting the computer to perform the right operations. They are able to do this because, as mentioned above, of people’s increasing demand for improved productivity and responsiveness of computers.

Next generation interface: Virtual Human’s potential

Computer machines that have been made with an emphasis on efficiency have a certain inorganic quality about them. In movies, “AI vs humans” is a commonly seen trope as well, and AI is often seen as simply a “tool” for humans. For AI in these cases, there is no warmth or kindness, nothing interesting about it.

That is why we are developing a next generation interface whose ability to communicate both “machine to human” and “machine to machine” is closer to that of a human being.

That is the Virtual Human.

In university experiments using the Virtual Human, the interface was better able to draw out personal information after directly interacting with the person rather than when just observing their interactions with other people.

In addition, the Virtual Human interface will become able to interact with computers through natural conversation, such as communicating with shop staff and receptionists. As a result, this means it could be used as one possible solution to the growing labor shortage as well as having many other meaningful impacts on society.

From the perspective of healthcare, research has shown that the Virtual Human’s ability to communicate can have many health benefits, such as helping reduce dementia and depressive symptoms and suppressing aging of the brain.

In this way, the Virtual Human’s greatest feature is its ability to communicate and interact with computers as a human interface. However, if conventional human customs and behaviors, and communication with respect to the other person’s facial expressions and movements are not observed ー if people feel something even slightly unnatural or out of place ー the interface will never be successful. Only once its behavior can be recognized as human-like will it ever come close to becoming a reality.

The Virtual Human is a major change from the one-sided system of humans commanding computers, and it therefore opens up a vast range of possibilities as a human interface. Because it has the ability to behave like a human this means that, just as each human is unique, the Virtual Human also has countless possible variations.

Making the Virtual Human a reality

Turning the Virtual Human into a reality is deeply entwined with the advancement of AI technology. First of all, in order to create a Virtual Human that behaves in a particular kind of way, a specific input is necessary ー just as human beings gather outside information from their senses and react accordingly. With the evolution of AI, characterized in particular by Deep Learning, it has become possible to distinguish inputs such as images, sounds, and natural language even more accurately than humans; it can be said that the foundation of the Virtual Human capable of using such inputs to communicate with humans has already been established. Now we must consider, just what kind of behavior would make the Virtual Human even more human-like?

Creating near-human AI engines through character design

Although the basis needed for creating human-like behaviors has definitely been established ー thanks to the progress seen in AI ー this does not necessarily mean it can exhibit super-human perception and recognition in a human-like manner. For example, smart speakers that can recognize human speech are currently able to properly interpret and convey those commands to a computer, but this is nothing more than just the ability to listen to human commands. No one can really say such interactions with smart speakers comes close to human-like.

In order to create more human-like behavior, another kind of AI is necessary. That AI is the one that has been used for decades for moving characters in the worlds of video games (“character AI”). Character AI does what humans naturally do ー that is, take in information from around them, process and remember it, and respond accordingly. Character AI uses this simple set of processes as a model for its own behavior. By using character AI’s ability to process and react to outside information, the Virtual Human gets ever closer to becoming more “human.” In our case, thorough research into these processes is reflected in the continuing development of our own AI engine. Furthermore, by incorporating character AI containing elements of human emotion, the Virtual Human becomes able to express various emotions too. As a result, it becomes possible to communicate even more naturally.

In order to create something capable of such human behaviors, we must first fully understand how humans are able to recognize, remember, process, and react to outside information, then use that as a model in the development of a character AI.

Next time, the focus will be on the process of typical human recognition and reaction that will form the base of this character AI.