A bold declaration was recently made by the biggest figure in generative artificial intelligence. Although he admitted he wasn’t being entirely serious…

Sam Altman, the CEO of OpenAI, may have been joking about “achieving” AGI (Artificial General Intelligence) but the comment still caused quite a stir, as the term has quickly built up a lot of significance in the months since ChatGPT was introduced to the public.

So what is Artificial General Intelligence? And why does it matter?

While it’s a broad and somewhat vague term, it has recently taken on the importance of being the assumed “win-state” of the race for AI dominance. Some have gone as far as to describe it as “the holy grail of computer science”.

The broadest but simplest definition of AGI is this…

An AI that matches human intelligence.

While it may be easy to argue that AI chatbots like ChatGPT and Bard have not only matched the reasoning powers of humans, they may have surpassed it.

Or at least some humans…

But these AI applications are still what is considered “narrow AI”. Or an artificial intelligence that is designed for a specific task and cannot apply what it has learned to a different task.

The main thing separating the smartest AI systems from even the (let’s say) least smart humans is that all humans apply lessons learned from past experiences to their future experiences, not just the experiences that are exactly the same.

According to Drew Sonden of SS&C Blue Prism says of AGI…

"A true AGI would be able to undertake conversational engagements, carry out detailed planning activities, deliver mathematical insights, create novel artworks -- in theory, undertake any task.”

While AGI is an aspiration for all of the major tech companies who have invested heavily in artificial intelligence, clearly steps have been taken to avoid taking such a step before the technology is truly ready.

ChatGPT, until very recently, was kept off of the world wide web as a precaution to keep it from totally uncontrolled learning.

All chatbots have some sort of review and refinement for their responses, manually ensuring that responses are accurate and also inoffensive.

A true AGI not only wouldn’t need these safeguards, they may not even be a possibility…

Crowning An AI Winner

While it’s not quite this simple, whoever creates an AGI first is assumed to be the ones who will be crowned the winner of the AI race. A trophy that will be valued at trillions of dollars, as AI is adopted by more and more industries.

But how will we know a “true” artificial general intelligence when we see it?

How can you tell if an AI is conscious or self-aware? After all, these are concepts that are difficult to define even within humans.

The most well-known test of a robot’s “human-ness” is the Turing Test (originally called the Imitation Game) and was introduced by the legendary mathematician Alan Turing in 1950.

While there is no officialset of questions for the Turing Test, the premise is to ask both a human and a machine a series of open-ended questions with no technically right answer, then have another human review the responses and try to determine which were made by the actual person.

These are some examples of questions that may be asked during a Turing Test…

• How would you describe yourself using only colors and shapes?
• What is the difference between a lie and a myth?
• What is the funniest joke you know?

Turing introduced this test with the words "I propose to consider the question, 'Can machines think?'".

This question had an answer, at least according to Turing…

No. Machines can not think.

But in 2014, the Russian chatbot named Eugene Goostman became the first machine to ever pass the Turing Test.

Eugene assumed the personality of a 13-year old Ukrainian boy and was able to fool a panel of 30 judges of varied backgrounds, including famed artificial intelligence professor Aaron Sloman.

But Eugene Goostman was not an AGI. It was an algorithm designed to pass the Turing Test and had no motivation outside of what it was programmed to do.

There is no official test to recognize an AGI. But an artificial intelligence that has its own motivations and broad reasoning skills will be hard to miss.

Not only would an AGI be able to efficiently solve the problems put before it, like narrow AI’s are already doing very efficiently, it would be able to identify problems that humans aren’t even aware of yet.

According to Chris Lloyd-Jones, the head of Open Innovation at Avande…

"If implemented with a view to our greatest challenges, [AGI] can bring pivotal advances in healthcare, improvements to how we address climate change, and developments in education”

While we use words like “fast” and “accurate” to describe AI now, once we see an AGI we’ll start to hear the word “brilliant”.

And whoever becomes the first one to present us with this holy grail will become the undisputed king of the biggest industry in the world.