Whether you’ve overheard a discussion concerning its philosophical tribulations or simply scrolled past a news article about some startup’s latest venture into the technology, AI seems to be the talk of the town nowadays.
But with AI being implemented everywhere from smartphones to hospitals, performing a seemingly endless number of different functions, it is important to understand the difference between levels of AI.
As stated, like any technology, AI has a wide range of applications and differing levels of complexity.
To account for this, AI is generally broken down into three broad categories: weak AI, artificial general intelligence (AGI), and the hypothesized artificial superintelligence.
If the names weren’t telling enough, weak, general and super denote the concerned AI’s technological sophistication, fitting it into one of the three categories.
Most people living in a developed nation have interacted with weak AI, and likely do so on a regular basis.
In fact, you can pick up your smartphone right now and interact with weak AI.
I do of course mean Siri, Cortana, Google Assistant and similar virtual assistants.
For example, Siri can obey simple voice commands such as “play Spotify” and, utilizing internet searches and apps, can answer simple questions such as “what is the capital city of France?” or “will it rain today?”.
Such virtual assistants can be classified as weak AI due to their limited forms of function and lack of ability to learn and apply new knowledge; they are simply a form of AI intended to make smartphones more user-friendly.
Most importantly, weak AI utterly lacks the capacity to problem solve or reason which are markedly human talents; largely leaving weak AI in the domain of handheld and household devices.
This is where artificial general intelligence enters the equation.
Artificial general intelligence
AGI is capable of reasoning, problem solving and learning at a level comparable to humans.
Impressively, AGI has beaten world-class Esports players and ancient board game champions at their own games, learning to outpace their human opponents with astonishing speed.
Increasingly it is also proving a boon to modern medicine, helping to diagnose medical conditions from rare neurodegenerative diseases to common flus.
Despite these feats, AGI is not conscious, nor is it capable of gaining consciousness through acquired knowledge.
A technological cornucopia, as of late 2017 there were 45 organizations actively researching AGI globally.
Artificial superintelligence, for better or worse, remains confined to the realms of science fiction and philosophical and practical debate.
Any machine possessing a level of intelligence clearly superior to that of humanity’s greatest minds in most or all disciplines would be categorized as artificial superintelligence.
It is not enough to outperform a human in only a video or board game, it would also have to outperform humans in knowledge and possibly application of science, mathematics, history and so on; an intellectual apex.
As artificial superintelligence is not currently a reality, there is no consensus on whether consciousness is a requisite.
Although without consciousness the machine would likely not be able to engage in legitimate philosophical or ethical discussion, philosophy and ethics are ideas that largely cannot be proven true or false, rendering outcompeting in these domains an impossibility.
If outcompeting humans in only quantifiable intellectual disciplines is the parameter for artificial superintelligence, then it is highly likely that it will eventually become a reality — assuming any rate of continued advancement.
Such an advanced form of intelligence will inevitably present practical and philosophical hurdles, but this is far from a reality, as the surface of AGI is only now being scratched.
Mitchell has a Bachelor of Arts with Majors in Journalism and Foreign Relations; and a Diploma of Digital Design.