In recent years the breakneck speed of exponential technological development has let many to believe that the realms of science fictions and human-like robots might become reality sooner than many think. I have been fascinated by these developments. Keen and curious to learn more about it delving down an exponentially expanding rabbit hole. Or a black hole so to speak. In this edition of the Book Club, we are going to explore the meaning behind the Singularity and its representation in “The Technological Singularity” by Murray Shanahan.
The Singularity in the context of the book and what most people refer to in recent years is the period in time in the future where a computer running software-based artificial intelligence reaches and surpasses that of humans. What this really means is that such technology and superintelligent agent can have an enormous impact on humanity and trigger time of recursive exponential development unfathomable to comprehend by a human being. A time that the mode of life and human existence change unrecognizably affected and dictated by superintelligence that will end life as we know it. The speed of this development has been speculatively measured by Ray Kurzweil who by implementing Moore’s Law, the law of accelerating returns and other mathematical and statistical methods came up with an approximation of this period falling around 2050. While this may be the predicted time period, how do we actually know when the AI software program has reached human intelligence. Artificial General Intelligence (AGI) or “strong AI” is the one capable of performing intelligent actions attributed to humans. This can be tested by passing the Turing test – “A machine and a human both converse sight unseen with a second human, who must evaluate which of the two is the machine, which passes the test if it can fool the evaluator a significant fraction of the time.” The foremost requirements for an AGI are common sense and creativity. The ability to reason and judge under uncertain circumstances. The ability to learn, plan and communicate on an ongoing basis in various different situations.
The book is written as an investigative piece of work without the timescale or prediction commitment. The aim of the book is not to make an exact prediction on when and where this might happen nor is it a piece of science fiction painting a futuristic story of a technological future ran by machines. Rather it gives a great overview of the subject matter and the possibility of its development. The book explores this development through 7 chapters. The first one, Routes to Artificial Intelligence discusses the different ways we can actually get to this point in the future, the technology required, its feasibility and possibility. The possibilities of technological singularity pose both existential risk and an existential opportunity and its initial development can set the tone for this. The second chapter explores one of those possibilities through a whole brain emulation of a human by mapping it then uploading it as a piece of software. The biological brain is a master of adaptation. Whole-brain emulation is a key to understanding the engineering behind human intelligence. Unlike the biological brain, a digitally realized brain emulation can be copied arbitrarily many times. And unlike a biological brain, a digital brain can be speeded up presenting unfathomable opportunities towards the path to developing superintelligence. The third chapter explores the creation of such super intelligence from scratch by trying to replicate and build the human brain on a piece of a software program then let it recursively improve itself. This seems hypothetically possible through a deep understanding of Neurology and Advanced Computer Science in the realm of Deep Learning, Neural Networks, Reinforcement Learning among other sub-Artificial Intelligence and Machine Learning methods implemented. Although there are certain concerns that we may reach the hardware limits of computing power and storage before this is actually is achieved. In the fourth chapter, a distinction is put between achieving human-level intelligence and surpassing it towards the development of superintelligence. Human intelligence is generalist first and adaptive then specialized. What is really hard for a machine is the exact opposite – being good at many different things rather than an expert in one single field. The fifth chapter of the book goes even further past the superintelligence exploring the notions of what makes us human and our consciousness how can this be
The last two chapters discuss the impact of the different levels of Artificial intelligence and how this can impact the lives of humans as they become obsolete in their daily life. Problems of higher purpose, drive and the meaning of life start surfacing as humans get pushed out of their vocations. Heaven or hell. Whatever happens, is completely unknown as much as is the motivation and actions of a superintelligent AGI. The Technological Singularity poses not only questions of physical capabilities of machines and their development but also philosophical questions on the essence of humanity and its fundamental value drivers. The author sees how the development of AGI can be achieved through continuous exponential improvement and expansion of knowledge in Big Data and AI. The most potent factor of the artificial intelligence development that is either based on whole brain emulation or reverse engineered from scratch would be its recursive self-improvement.
Intellectual prowess is what sets human beings apart from animals. Can we teach machines the same prowess to distinguish and elevate them above humans? That is yet to be discovered.
“What then am I? In the end, all we have is simply what we find, and what we can usefully say to each other about what we find is all that needs to be said. And perhaps, in the end, it’s best just to sit quietly and let go of that thought too.”