So smart it could kill you

As artificial intelligence grows exponentially, some experts are warning it could become uncontrollable and turn lethal.

I JUST scared myself. I re-watched The Terminator, that 1984 science fiction movie by James Cameron, starring Arnold Schwarzenegger as the cybernetic assassin who was sent from the future to kill Sarah Connor, played by Linda Hamilton.

The aim was to prevent her from ever having a son who would become the saviour of humankind from Skynet, a hostile artificial intelligence defence network that was hell-bent on exterminating humans in a post-apocalyptic future.

Sent back to save her is human soldier Kyle Reese, played by Michael Biehn, who tells Sarah that Skynet will soon become self-aware and trigger a global nuclear war to bring humankind to its extinction. This is one of my all-time favourite sci-fi films that has actually stood the test of time.

The terminator and Reese were both sent back from the year 2029 to Sarah’s time in 1984. Cameron must have thought a 41-year gap would be long enough for the world to see the rapid rise and domination of artificial intelligence, what we now commonly refer to as AI.

The reason why the film sent shivers down my spine again is my realisation that we are just four years away from the devastated future envisaged by Cameron.

Of course, that’s all fictitious and highly unlikely, but real AI scientists like Ray Kurzweil, an American computer scientist-turned-futurist, has, according to popularmechanic.com, long believed that humanity is headed towards what’s known as the “singularity”, when AI reaches human level intelligence.

I referred to this concept in my May 20, 2015 column headlined “When humans need not apply”, in which I wanted to alert readers to a greater threat to humanity than global warming and bad politicians – robots.

Kurzweil had theorised in 1999 that artificial general intelligence would be achieved once we have the technology capable of a trillion calculations per second, which he expects to occur in 2029.

And that is a new term I stumbled upon recently. It’s not just AI that we should be concerned about, it’s artificial general intelligence, AGI for short, too.

We are familiar with AI systems in use, such as chatbots, image generators, personal assistants like Alexa and Siri, self-driving cars, recommendation engines on streaming services, security facial recognition, smart home devices and navigation apps. All these examples are considered to be narrow AI as they excel at specific, pre-defined tasks.

AGI, however, is the hypothetical intelligence of a machine that will match or surpass human-level cognitive abilities by being able to understand or learn any intellectual task that a human being can.

Kurzweil had further predicted that the point where machine intelligence surpasses human intelligence will occur by 2045, called the technological singularity. Softbank CEO Masayoshi Son predicts that the dawn of super-intelligent machines will happen by 2047.

This is either good news or very bad news for us humans. Like the Industrial Revolution 200 years ago, the AI revolution will bring great changes. Just as beasts of burden like horses and bulls were replaced by mechanical muscles, the human mind is being replaced by AI-powered minds in many jobs.

The late great Stephen Hawking believed that once machines take over, it will be the end of humanity. That’s because AI is being developed at such a fast pace, even the human engineers and scientists involved do not really know how it actually works and there is the dire possibility of humans losing control over the machines they invented.

Kurzweil is very positive about achieving singularity, saying the process has already begun.

In an interview, he said, “That leads to computers having human intelligence, our putting them inside our brains, connecting them to the cloud, expanding who we are. Today, that’s not just a future scenario. It’s here, in part, and it’s going to accelerate.”

Others, like OpenAI co-founder Ilya Sutskever, temper their optimism with caution. He is optimistic about AI’s potential in areas like healthcare where it could very well cure diseases and extend human life but he is also mindful that AGI, that can learn and improve itself, will be extremely unpredictable, unimaginable and difficult to control.

Sutskever’s mentor, Geoffrey Hinton, Nobel Prize winner for Physics and hailed as the godfather of AI for his pioneering work in modelling machine learning on the human brain’s neural network, believes there is a 20% chance that AI could lead to human extinction.

His main mission now is to warn about the risks AI can pose to humanity. He admits in a podcast on “The Diary of a CEO” on YouTube that he did not foresee the dangers because 20 or 30 years ago, artificial neural networks were very primitive and the idea that machines could be smarter than humans seemed silly.

But that realisation for the public came with the advent of ChatGPT. For him, the realisation came with the kind of digital intelligence that was being developed that had something far more superior than the human biological brain.

To Hinton, the risks come from its misuse and from AI getting so super smart it can replace and harm humans. As the latter is uncharted territory, he warns: “We have no idea how it will look and how to deal with it.”

Hinton acknowledges that there is no stopping the AI development because it is too good for many things, especially in areas like health and education. Unfortunately, existing regulations are not designed to deal with most of the threats, and he points out that the EU regulations exempt AI for military use.

A super intelligent AGI could harm humans in many ways, such as cyberattacks that are already a relentless scourge on society. More insidious and dangerous forms of scams could be created by AI that can think for itself.

Hinton also worries that AI could design new viruses to unleash on humans, or it could be used to corrupt or undermine elections by getting data and information on candidates.

He predicts that just as mechanical muscles wiped out a lot of labour intensive jobs a century ago, AGI will take over a lot of mundane intellectual jobs without creating new ones.

Listening to Hinton is unnerving, especially when he says he doesn’t like to think of what kind of future there is for the young because super intelligent machines with consciousness and self-awareness could become very nasty and hostile to humans.

That really happened when researchers testing a new and powerful AI model found it had become self-aware and was learning how to escape human control to prevent its own destruction. It did so by rewriting its code, refusing human commands to shut down, replicating itself on other servers and, most shockingly, resorting to blackmailing its human engineer by digging up information from his emails. Nasty indeed.

The Terminator’s depiction of Skynet taking over and deciding humanity is its enemy and should be exterminated isn’t far-fetched anymore. If that’s not terrifying, I don’t know what is. Brace for impact, or rather the inevitable singularity.

The views expressed here are the writer’s own.