Can machines think?
Humans have pondered this question since the times of the ancient Greeks, and the Myth of bronze automaton Talos — forged by the god Hephaestus to protect the island of Crete — says Ray Kurzweil in his timely and mind-bending new book, “The Singularity is Nearer: When We Merge with AI.”
But it wasn’t until 20th-century computer scientist Alan Turing came along that this philosophical question became something we could test empirically.
Turing’s “imitation game” — now known as the Turing Test — sought to determine whether a machine could perform the same tasks as a human brain.
If the machine successfully fooled the tester into believing they were speaking to a person — rather than a computer — the machine was said to have “passed the test.”
In the decades since, the Turing Test has acted as a North Star for computer programmers who have tried, and largely failed, to build machines that could pass the test — until now.
Building off of his seminal 2005 work, “The Singularity is Near,” which first explored the potential for man and machine to merge – Kurzweil’s latest argues, convincingly, that artificial intelligence has already crossed the Rubicon.
He writes that AI is now exceeding the human brain at several cognitive tasks and that it will eventually do all things far better than even the most expert humans.
These new machines can learn, reason, plan and act with intention, and they are becoming far smarter far faster than most people, save Kurzweil, could have predicted.
Soon, he forecasts, they will be indistinguishable from human brains, before accelerating past them in nearly every way.
Things are moving so quickly, in fact, that we have become desensitized to the pace of progress. “Before AI achieves some goal, that goal seems extremely difficult and singularly human. But after AI reaches it, the accomplishment diminishes in our human eyes.”
Kurzweil’s boldest idea is that AI is ushering in a new epoch where biological intelligence and computers will quite literally merge.
This event he calls the Singularity, which he sees happening around 2045.
For Kurzweil, the stakes of this epochal shift are high.
Our very survival is in question if we do not manage it effectively.
“AI technologies are inherently dual use” meaning they can be wielded for good or bad.
Kurzweil believes we should embrace this future and not fear it, and celebrate AI, while preparing ourselves for profound disruptions to the economy, society, our biology and our way of life.
In Kurzweil’s telling, in the not-too-distant future, we will be able to augment our brains to become vastly more intelligent and extend our life span to 1,000 years and beyond.
“We will not be dependent on the survival of any of our bodies for our selves to survive,” he says.
We will choose to work for spiritual fulfillment, not to make a living.
For decades, many of Kurzweil’s many prescient predictions — seemingly outlandish in the time — have come true.
AI is the latest effort in our “long quest to reinvent our own intelligence” and a clear example of what Kurzweil calls the “law of accelerating returns, where information technology creates feedback loops of innovation.”
Not all innovation is exponential, he says.
Passenger airliners in the 1960s traveled about as fast as today’s jets.
But today’s most basic smartphone, which cost around $50, would trounce any supercomputer from the 1960s.
And things are not slowing down: “Computations used to train a state-of-the-art artificial intelligence model has been doubling every 5.7 months since 2010,” outpacing Moore’s Law which says performance should roughly double every 18 months.
Kurzweil does his best work comparing human brains to AI computers known as “neural nets” to reveal starting similarities.
Whereas traditional computers perform operations sequentially, AI models work like human brains and complete multiple operations in parallel.
Traditional computers tried to achieve artificial intelligence by using ‘brute force’ and applying as much computing power as possible.
Today’s AI models break down language into chunks, or “tokens,” and then rearrange them, learning meaning “not from a grammatical rule book or dictionary but rather from the contexts that words are actually used In.”
And, like people, Modern AI is also “multimodal,” meaning it can answer queries in plain language but also create still images, video and even write computer code too.
Kurzweil says it will be hard to say when exactly AI will have surpassed humans.
“It might turn out, for example, that in 2034 AI can compose Grammy-winnings songs but not write Oscar-winning screenplays, and can solve Millennium Prize problems in math but not generate deep new philosophical insights,” he said.
The book contains several real-world examples of how AI is already impacting many industries, surpassing human capability, specifically in healthcare.
AI models can recognize patterns in huge datasets, leading to more accurate diagnosis of disease.
AI simulations can complement clinical trials, reducing development cost and improving the efficacy of new drugs. Indeed, the FDA is now incorporating simulation results in regulatory approvals.
Robotic “surgeons” can be trained on millions of surgeries and simulations, far more than a human surgeon could ever do in a single lifetime, while AI models can simulate new proteins that can unlock the mysteries of human biology.
Kurzweil credits AI with sequencing the coronavirus and accelerating the development of mRNA vaccines, saving countless millions of lives.
The book also explores what AI will mean for jobs and the economy.
Kurzweil acknowledges the potential for disruption: 2.7% of workers today serve as some kind of driver, and in some states, like New York and New Jersey, it is as high as 8%.
“Yet driving is just one of a very long list of occupations that are threatened in the fairly near term by AI,” he says. “Other jobs are also at risk, such as telemarketers, insurance underwriters, and tax preparers.”
Kurzweil believes there will be a period of disruption, followed by an era of plenty and that “we would effectively have universal basic income (UBI) or its equivalent.”
In this future “our main struggle will be for purpose and meaning,” he writes.
Kurzweil’s book reminds us that “the future is here, it’s just not evenly distributed.”
We already live in a society with vastly different outcomes based on factors like wealth and geography. Could AI exacerbate them, at least in the short term?
What happens to society if the life expectancy of low-income Americans flatlines while it soars by hundreds of years for the rich who can access life-extending and brain-augmenting technologies?
Kurzweil worries about “toxic politics” that could “interfere with rising living standards,” but do we even have competent politicians who are thoughtfully considering these issues?
“The Singularity is Nearer” is a feast for the (as-of-yet unenhanced) mind.
It educates and illuminates and asks profound questions that get to the heart of our relationship to technology and our own sense of “humanness.”
Can a computer with a brain indistinguishable from a human’s feel emotion like envy, anguish, or joy?
Do they deserve protection under the law, even human rights?
At what point does a machine gain sentience and will we ever really know it?
If a person’s entire consciousness can be downloaded to a computer, is it still a human mind?
At that point, are we still human?
Readers will be left pondering these questions long after they’ve closed the book.
