Reading more like a hyper extension of Aldous Huxley’s “Brave New World” than an introduction to an explosion of technology, James Barrat’s thought provoking book, “Our Final Invention”, brings us face to face with the alarmingly potential consequences of Artificial Intelligence (“AI”) gone uncontrolled. Warning the reader of an impending doom, Barrat finishes his book on a dystopian note by writing that while human-level AI will have countless applications, it will “fundamentally alter our existence (if it doesn’t end it first)”.
Martin Ford, a futurist and author focusing on artificial intelligence and robotics, and the best selling author of “The Rise of the Robots”, in an upcoming work, appropriately titled “The Rule of The Robots”, explores in a measured, methodical and meticulous manner the inevitable status of ubiquity that AI will attain in the future. While conceding that there is no escaping the fact that there would be a proliferation of AI in the coming days and decades, Ford sticks his neck out to proclaim that AI will ultimately evolve into a ‘general-purpose technology of such scale and power that it can be compared to electricity.’
Extraordinary advances in branches such as deep neural networks, and machine learning have opened up a plethora of possibilities hitherto unimagined. The Holy Grail underpinning all these complex research is to birth an Artificial General Intelligence or an Artificial Super Intelligence even that would mimic the human brain. Once this phenomenon manifests, the rate of knowledge expansion would be exponential rather than incremental. The AI would reach a trajectory of ‘unsupervised’ learning and thereby leave mankind gasping when it comes to super intelligence. This is also the pinnacle where, according to experts such as the late Stephen Hawking, entrepreneur Elon Musk, MIT physicist Max Tegmark and intellectuals such as Sam Harris etc. assurance ends, and anxiety begins. Companies such as Intel and IBM have already made significant progress with “neuromorphic” chip designs. These designs instantiate hardware versions of the neurons in human brain, directly onto silicon. The Loihi chip of Intel encompasses 130,000 hardware neurons each of which in turn connects to a multitude of other similar neurons.
As Ford illustrates in an exhilarating manner, the field of deep learning has revolutionised AI and the output emanating from this sphere has almost blurred the line separating reality from science fiction. In the year 2019, the Turing Award (popularly known as the “Nobel Prize of Computing”) was bagged by Geoffrey Hinton, Yann LeCun and Yoshua Bengio. The trio won this distinction for their unparalleled and original contributions in the filed of deep learning. Ford provides an invaluable primer on the science and technology behind deep learning. Deep learning employing an artificial neural network attempts to create a rough numerical blueprint of the way neurons in the human brains operate and interact. Upon configuration such a network is trained to accomplish specific tasks such as image recognition and language translation. Every ‘layer’ of the network enhances the level of abstraction of the knowledge fed to the network. Deep learning’s apex achievement came in the year 2016 when AlphaGo, a system developed by DeepMind defeated one of the greatest “Go” boardgame players Lee Seedol in a five-match game in Seoul. This almost represented an inflection point in the field of AI. Alpha Go had bested a champion in a game whose very intricate permutations and combinations almost made it immune to brute force algorithms.
However, as Ford writes, mankind is still way behind realizing the true potential of Artificial General Intelligence. Ford interviewed twenty three pioneering researchers in the field of AI for his book “Architects of Intelligence.” Many of the researchers expressed their strong skepticism about the advent of superintelligence. For example Andrew Ng, who was engaged in AI research at companies such as Google and Baidu, said that “worrying about an existential threat from AI is like worrying about overpopulation on Mars – long before even the first team of astronauts has been sent to the red planet.”
The experience and experiment with ‘self-driving’ or ‘automated’ cars lends a critical and seminal insight into both the progress and pitfalls of AI. Founder of the autonomous driving startup Aurora and also the former chief technology officer of Google, Chris Urmson once brazenly declared in 2015 that his then eleven year old son would have a redundant driving license when he turned sixteen. But Urmson’s declaration has turned out to be a mere speculation as the world is still trying to grapple with many an intricacy and nuance surrounding the technology underlying a self-driving car. Stefan Seltz-Axmacher, the CEO and co-founder of Starsky Robotics, avers that a critical factor throwing sand in the gears of the self-driving automobile industry is an inherent lacuna of deep learning. Paraphrasing Ford, “a system with the flexibility to truly offer autonomous driving under all circumstances without the need for remote human supervision, may well be beyond the capability of today’s deep learning system and is unlikely to arrive in the near future.”
However, as Ford illustrates, the market for AI is burgeoning with inventive and innovative applications. Powered by a funding of over $200 million from investors such as Mark Zuckerberg, Samsung, Elon Musk and Jeff Bezos, technology company Vicarious is revolutionizing a “robots as a service” business. Acquiring industrial robots from various companies, Vicarious integrates the robots with AI software conceptualised by its own team before proceeding to rent the robots out to whoever needs them.
Ford also touches upon a very topical conundrum associated with untrammeled advances in the field of AI. In a world characterized by rampant technology, the lines between routine and non-routine stand blurred. Both routine as well as non-routine tasks that are emblematic of repetition are under the threat of being obliterated by AI. The former Bank of England Governor and currently, Vice Chairman and Head of Impact Investing at Brookfield Asset Management, Mark Carney to exclaim that it was time for a “massacre of the Dilberts.” Machines of the modern era are not just capable of performing routine activities, but as AlphaGo illustrated in searing detail, are also extremely capable of executing tasks that require cognitive skills and affectations, traits that are ‘non-routine’ by all stretches of imagination. Netscape founder and world renowned venture capitalist Marc Andreessen immortally said, “software is eating the world” and that in the end, there will be only two types of people left: those who program the machines, and everyone else. According to Ford, the ongoing Coronavirus pandemic has also worsened the grim prospects of unemployment. “Indeed, the economists Nir Jaimovich and Henry E. Siu studied this phenomenon and in a 2018 paper found that ‘essentially all employment loss in routine occupations occurs in economic downturns’” Businesses reeling under economic strains are forced to furlough or lay off workers. At the same time with a view to extracting the maximum yields and efficiencies, these companies also make increasing use of technology such as automation. Once stabilization rears its head and the economy gets back to normalcy, the companies do not find any reason to hire back the employees that they have let go.
Ford also mulls about the invasive nature of AI by making a chilling reference to the sophisticated surveillance ecosystem perfected by China as a result of which more than 1.8 million Uighurs find themselves incarcerated in draconian concentration camps, euphemistically named “reeducation centres”. Other applications of AI such as drone technologies and ‘deepfakes’ provide reasons to be apprehensive. A development known as “generative adversarial network” (GANs) has exacerbated the already dangerous power of deep fakes.
So what are these GANs? GANs represent algorithmic architectures employing two neural networks. These two neural networks are pitted against each other. The quintessential purpose behind such a face off being to generate new, synthetic instances of data that can pass for real data. GANs are liberally used to generate images videos and voice. The man behind the propagation of GANs was Ian Goodfellow, a researcher at the University of Montreal. While GANs have immense value in the domains of music and speech, their potential to birth evil is also immense. GANs are the primary vehicles to purvey Deepfakes by using voice and image overlays for derogatory and depraved purposes.
Ford ends his book in a deeply poignant and experiential manner. Envisioning the AI future as a continuum, Ford has two diametrically different scenarios at either end of the continuum. At one end lies Star Trek. This immortal television series featuring the endearing spaceship ‘Enterprise’ with the debonair Captain James Kirk at the helm. The world of Star Trek is one filled with prosperity, abundance, fulfillment and shorn of poverty and skull drudgery. There is a total paucity or absence of conventional jobs, yet people live a life of contemplation, exploration and introspection. There is an appreciation of the values attributed to intrinsic humanity rather than monotonous output.
At the other extreme end of the continuum lies the bleak and terribly dark movie ‘The Matrix’. However unlike the picture where a world enslaved by AI looks to an unlikely saviour to extricate itself from the vice-like grip of its tormentor, Ford’s Matrix has a ‘balkanized’ society where a segment of the privileged elite remain anchored in the actual world, while the rest of the teeming masses find themselves absorbed in an addictive world of technology or the murky and dark bowels of crime.
It is ultimately in the hands of the world to choose its future.
(Rule of the Robots is published by Perseus Books and Basic Books and will be released on 14 September, 2021)
The reviewer thanks NetGalley for the Advanced Reviewer Copy