What Is the AI Singularity, and Is It Real?

June 2024 ยท 8 minute read

Key Takeaways

As AI continues to advance, the topic of the singularity becomes ever more prominent. But what exactly is the singularity, when is it expected to arrive, and what risks does it pose to humanity?

What Is the AI Singularity?

Sci-fi films have toyed with the idea of the singularity and super-intelligent AI for decades, as it's a pretty alluring topic. But it's important to know before we delve into the details of the singularity that this is an entirely theoretical concept at the moment. Yes, AI is always being improved upon, but the singularity is a far-off caliber of AI that may never be reached.

This is because the AI singularity refers to the point at which AI intelligence surpasses human intelligence. According to an Oxford Academic article, this would mean that computers are "intelligent enough to copy themselves to outnumber us and improve themselves to out-think us."

As said by Vernor Vinge, the creation of "superhuman intelligence" and "human equivalence in a machine" are what will likely lead to the singularity becoming a reality. But the term "AI singularity" also covers another possibility, and that's the point at which computers can get smarter and develop without the need for human input. In short, AI technology will be out of our control.

While the AI singularity has been posed as something that will bring machines with superhuman intelligence, there are other possibilities, too. A level of exceptional intelligence would still need to be reached by machines, but this intelligence may not necessarily be a simulation of human thinking. In fact, the singularity could be caused by a super-intelligent machine, or group of machines, that think and function in a way that we've never seen before. Until the singularity occurs, there's no knowing what exact form such intelligent systems will take.

With network technology being invaluable to how the modern world works, the achievement of the singularity may be followed by super-intelligent computers communicating with each other without human facilitation. The term "technological singularity" has many overlaps with the more niche "AI singularity", as both involve super-intelligent AI and the uncontrollable growth of intelligent machines. The technological singularity is more of an umbrella term for the eventual uncontrollable growth of computers, and also tends to require the involvement of highly intelligent AI.

A key part of what the AI singularity will bring is an uncontrollable and exponential uptick in technological growth. Once technology is intelligent enough to learn and develop on its own and reaches the singularity, progress and expansion will be made rapidly, and this steep growth won't be controllable by humans.

In a Tech Target article, this other element of the singularity is described as the point at which "technology growth is out of control and irreversible." So, there are two factors at play here: super-intelligent technology, and the uncontrolled growth of it.

When Is the Singularity Expected?

To develop a computer system capable of meeting and exceeding the human mind's abilities requires several major scientific and engineering leaps before it becomes a reality. Tools like the ChatGPT chatbot and DALL-E image generator are impressive, but I don't think they're anywhere near intelligent enough to earn singularity status. Things like sentience, understanding nuance and context, knowing if what's being said is true, and interpreting emotions, are all beyond current AI systems' capabilities. Because of this, these AI tools aren't considered to be intelligent, be it in a human- or non-human-simulated fashion.

While some professionals think that even current AI models, such as Google's LaMDA, could be sentient, there are a lot of mixed opinions on this topic. A LaMDA engineer was even placed on administrative leave for claiming that LaMDA could be sentient. The engineer in question, Blake Lemoine, stated in an X post that his opinions on sentience were based on his religious beliefs.

LaMDA is yet to be officially described as sentient, and the same goes for any other AI system.

No one can see the future, so there are many differing predictions regarding the singularity. In fact, some believe that the singularity will never be reached. Let's get into these varying viewpoints.

A popular singularity prediction is that of Ray Kurzweil, the Director of Engineering at Google. In Kurzweil's 2005 book, 'The Singularity Is Near: When Humans Transcend Biology', he predicts that machines that surpass human intelligence will be created by 2029. Moreover, Kurzweil believes that humans and computers will merge by 2045, which is what Kurzweil believes to be the singularity.

Another similar prediction was posed by Ben Goertzel, CEO of SingularityNET. Goertzel predicted in a 2023 Decrypt interview that he expects the singularity to be achieved in less than a decade. Futurist and SoftBank CEO Masayoshi Son believes we'll reach the singularity later on, but possibly as soon as 2047.

But others aren't so sure. In fact, some believe that limits on computing power are a major factor that will prevent us from ever reaching the singularity. The co-founder of AI-neuroscience venture Numenta, Jeff Hawkins, has stated that he believes "in the end there are limits to how big and fast computers can run." Furthermore, Hawkins states that:

We will build machines that are more 'intelligent' than humans, and this might happen quickly, but there will be no singularity, no runaway growth in intelligence.

Others believe the sheer complexity of human intelligence will be a major barrier here. Computer modeling expert Douglas Hoftstadter believes that "life and intelligence are far more complex than the current singularitarians seem to believe, so I doubt it will happen in the next couple of centuries."

Why Are People Worried About the Singularity?

Humans have lived comfortably as the (as far as we believe) most intelligent beings in known existence for hundreds of thousands of years. So, it's natural for the idea of a computer super-intelligence to make us a little uncomfortable. But what are the main concerns here?

The biggest perceived risk of the singularity is humanity's loss of control of super-intelligent technology. At the moment, AI systems are controlled by their developers. For instance, ChatGPT can't simply decide that it wants to learn more or start providing users with prohibited content. Its functions are defined by OpenAI, the chatbot's creator, because ChatGPT doesn't have the capacity to consider breaking the rules. ChatGPT can make decisions, but only based on its defined parameters and training data, nothing further. Yes, the chatbot can experience AI hallucination and unknowingly lie, but this isn't the same as making the decision to lie.

But what if ChatGPT became so intelligent that it could think for itself?

If ChatGPT became intelligent enough to dismiss its parameters, it could respond to prompts in any way it wants. Of course, significant human work would need to be done to bring ChatGPT to this level, but if that ever did happen, it would be very dangerous. With a huge stock of training data, the ability to write code, and access to the internet, a super-intelligent ChatGPT could quickly become uncontrollable.

While ChatGPT may never achieve super-intelligence, there are plenty of other AI systems out there that could, some of which probably don't even exist yet. These systems could cause an array of issues if they surpass human intelligence, including:

According to Jack Kelley writing for Forbes, AI is already causing job displacement. In the article, job cuts at IBM and Chegg are discussed, and a World Economics study about the future of the job market with AI is also included. In this report, it is predicted that 25 percent of jobs will be negatively impacted over the next five years. In the same study, it was stated that 75 percent of global companies are looking to adopt AI technologies in some way. With this huge proportion of the worldwide industry taking on AI tech, job displacement due to AI may continue to worsen.

The continued adoption of AI systems also poses a threat to our planet. Powering a highly intelligent computer, such as a generative AI machine, would require large amounts of resources. In a Cornell University study, it was estimated that to train one large language model is equal to around 300,000 kg of carbon dioxide emissions. If super advanced AI becomes a key part of human civilization, our environment may suffer considerably.

The initiation of conflict by super-intelligent AI machines may also pose a threat, as well as how machines surpassing human intelligence will affect the global economy. But it's important to remember that each of these pointers is dependent on the AI singularity even being achieved, and there's no knowing if that will ever happen.

The Singularity May Always Be a Sci-Fi Notion

While the continued advancement of AI may hint that we're headed towards the AI singularity, no one knows if this technological milestone is realistic. While achieving the singularity isn't impossible, it's worth noting that we have many more steps to take before we even come close to it. So, don't worry about the threats of the singularity just yet. After all, it may never arrive!

ncG1vNJzZmivp6x7qbvWraagnZWge6S7zGiuoZmkYra0edOhnGaZmWLAqrrGrqOaqpmpxm6tzZ1koqtdnsFuvsSao2g%3D