Artificial General Intelligence, broadly speaking, is a distinguished type of AI that enables machines to carry out tasks that humans are capable of, and also those that most humans aren’t capable of.
The term is relative to what is commonly known as Artificial Intelligence (also known as Weak Intelligence) which only enables machines to perform specific tasks. Artificial General Intelligence (contrastingly known as Strong AI), however, enables systems to find a solution and successfully tackle even unfamiliar tasks.
At the core of the theory of Artificial General Intelligence, lies the aim to enable machines to mirror the human cognitive systems or the capabilities of the human brain. To what extent is this possible, and if it is at all fully possible, how far are we from realizing this objective is what we will be diving into.
The concept of Artificial General Intelligence is best illustrated by the book ‘On Intelligence’ by Jeff Hawkins. In his book, he uses the Chinese Room thought experiment to cleverly illustrate the difference between AI and AGI.
The thought experiment goes something like this – Imagine a person in a room who does not understand Chinese (this person acts as the proxy) but has a very large rule book. Given a sheet of paper with a Chinese sentence the person/machine uses the rule book to translate it into English and gives us an output. Further imagine the rule book is so good that this arrangement passes the Turing test ie it convinces a native Chinese speaker that she/he is in fact talking to a native speaker of Chinese. The question posed is – Does the person/machine truly UNDERSDTAND Chinese? Or are they just SIMULATING the ability to understand Chinese?
This core thought experiment probably marks the divide between AI and AGI. While AI algorithms are getting very good at specific tasks they are trained on – It is still questionable whether they understand what exactly they are doing.
But before we establish the possibility of machines realizing human capabilities better than we do ourselves, let’s take a quick detour into understanding where we are today, with Artificial Intelligence and how heavily invested an area of science it is.
Where is Artificial Intelligence today?
“Throughout human history, we have been dependent on machines. Fate, it seems, is not without a sense of irony.”
Thus said Morpheus to Neo in the 1999 film, The Matrix.
We could relegate the validity of the statement as merely a powerful dialogue from a science-fiction movie.
But with the amount of research and the kind of dollars that countries across the world are investing in AI technology, it might not be all that wise to disregard it!
According to this report by IDC that was published in March 2019, worldwide spending on Artificial Intelligence (AI) Systems is said to reach $35.8 billion in 2019- a 44% jump from what was spent in 2018. IDC also expects that spending on AI systems will more than double to $79.2 billion by 2022, and to $97.9 billion by 2023!
If this wasn’t enough, here are a few other stats that will establish the fact that AI systems are one of the most heavily invested areas, globally:
- The city of Beijing alone is investing $2 billion dollars in AI, keeping up with China’s declaration of being the world leader in AI by 2030. (Source)
- According to this report by Deloitte, France released its “AI for Humanity” initiative in 2018 dedicating €1.5 billion towards enacting the plan.
- The same report also shows that Venture Capital Investment alone in the United States stands at $8 million, as of 2018.
- The US government, on the other hand, has spent almost a billion dollars in AI-centric research grants, since 2003. And, more than half of this amount has only been awarded post 2015. (Source)
- Israeli AI startups saw a fifteen- fold increase in the money raised in 2017, when compared to the 5 preceding years, which stood at $837 million. (Source)
These are indicators that from Tel Aviv to Beijing to San Francisco, several forces are working towards making Artificial Intelligence more efficient and powerful. Which leads us to the inevitable realization that perhaps Artificial General Intelligence is possible, after all!
How do we, then, achieve Artificial General Intelligence?
Businesses from all industry types have tasted success with AGI. Whether it is healthcare, retail, banking, aviation, software, academia, or even travel & hospitality, the idea of going back to a world without AI seems unimaginable. With such profound dependency on AI technology, it is but obvious that this proverbial golden goose will only be explored further, if not exploited.
But for AI technology to leap to Artificial General Intelligence is no joke. Yes, we have self-driven cars, the world’s first AI legal assistant ROSS, virtual AI-powered assistants SIRI, Cortana and the likes which can process and understand text and speech delivered in natural language. They may be incredibly perceptive when it comes to understanding and responding to humans, but they are still performing very specific tasks. They abide by innumerable rules of ‘if, then’ and store insane amounts of data. But that’s about it!
The concept of Artificial General Intelligence encompasses more than just NLP or retaining large amounts of data. Performing a specific task better than humans is not the same as being “intelligent” and that’s where the challenge really lies.
Achieving Artificial General Intelligence will be possible when machines are capable of some human capabilities, like:
- Learning with fewer experiences rather than from incredibly large amounts of training data
- Applying learnings from one experience to another or others in general (Transfer learning)
- The ability to understand and differentiate between various ‘contexts’ as well as humans do
- The power to use NLP of one language to certain others that form part of the same linguistic family making it that much more human-like
So how far are we from engineering AI technology that thinks like humans?
Writer and Futurist Martin Ford interviewed 23 of the biggest names in the Artificial Intelligence space for his book “Architects of Intelligence”. The interviewees were some of the world’s foremost researchers and entrepreneurs who had dedicated a significant part of their lives to understanding and developing AI technology.
As part of the discussions, he informally asked each of them to guess by when they foresee at least a 50% chance of AGI being realized. Of the 23, only 18 responded to the question and of those, only 2 went on record with their answers. Futurist and Director of Engineering at Google, Ray Kurzweil, believed that this could happen by 2029. Whereas Roboticist and Co-Founder of iRobot, Rodney Brooks, suggested that it could well be a while, mentioning that 2200 is more like the year when Artificial General Intelligence could see at least a 50% chance of being realized.
With that much disagreement between some of the most elite minds from within the AI ecosystem, it is but evident that is a question that is nowhere near ready to be answered, let alone brought to life.
Responsible AI & the Frankenstein effect:
Writer Mary Shelley published the book “Modern Prometheus”, popularly known as “Frankenstein” in the year 1818, when she was still a teenager. This was a time when AI, as we now know it, did not exist. But several critics have drawn similarities between the novel and the potential threat posed by AI technology, especially when pushed to outpace itself by megalomaniacal entrepreneurs and scientists.
It is, of course, a valid fear that even Stephen Hawking, one of the world’s most renowned scientists, harbored.
In an interview he gave to BBC in 2014, Hawking said, “The development of full artificial intelligence could spell the end of the human race. It would take off on its own, and re-design itself at an ever-increasing rate. Humans, who are limited by slow biological evolution, couldn’t compete, and would be superseded.”
Future of AI:
So, with uncertainty being the only certainty for now, I’m but forced to leave you with another of Morpheus’ quote from ‘The Matrix’.
“Unfortunately, no one can be told what the matrix is. You have to see it for yourself.”
And I suppose we will only have to experience the future to truly understand the future of AI – where Artificial Intelligence goes and takes us, how it defines the course of humanity and if a machine is capable of being more human than us humans.