The Moral Imperative of Artificial Intelligence
Historically, human society has evolved by turning the wheels, and adding the axles, before tightening the grip on the steering wheel.
The lurking perils of artificial intelligence
While artificial intelligence (AI) offers definitive benefits, it behooves us to consider the unintended negative consequences of human innovations. AI is a dual-use technology that can be repurposed for nefarious use by state actors and digital crime syndicates. The line between deep-learning and deep-fake is treacherously thin. There may be numerical probabilities and likelihoods in mathematics, but there also are immense capabilities and sustainable livelihoods in people. The commercialization, rather than democratization, of AI will lead to data capitalism, widen local and global inequalities, and the emotional and psychological manipulation of people. A power shift to digital colonies may deepen the inequity between the advantaged and the disadvantaged, rather than value creation for the betterment of society as a whole. This may inevitably trigger massive labor displacement and major mid-career job retraining for the displaced workforce. The weaponization of AI to usher in an era of hyper-scale automation and efficiencies may lead to de-population and centralization of data banks amongst the corporate elites. AI based manipulation of human minds and societal behaviors, using personal and group information sniffed overtly or covertly from data traffickers, to exploit emotional and psychological comfort zones with sentiment analysis poses serious risks to humanity. There may be fuzzy logic in mathematics and economics, but there is nothing fuzzy about the hopes and aspirations of people. No civilization in the history of mankind has been immune to the innocuous forces of radical shifts that trigger the collapse of societies. Unlike the migration to mobile technologies and consumer devices over the past decades, the future of AI holds an immense and grave potential to deconstruct, reduce, and manipulate individuals and groups, replace the truth with popularism, and label facts as opinions.
Where human minds and machines diverge
Programming machines with human minds, to re-program human minds with pre-trained machines, may become a catastrophic experiment in the manipulation of the human psyche. Moore’s law applies only to computers and not to the human brain. Generation Z do not possess more memory relative to generation Y, X, baby boomers, or even the lost generations. The elastic compute power of silicon does not apply to the human species. A computer’s resources (memory, compute, and storage) can scale-up elastically, whereas the capacity of the human mind diminishes in old age. While aging machines become obsolete over time, they continue to perform at full capacity with regular maintenance and updates. The aging process of humans gradually lowers the performance of both their body and mind. The mind of a machine is superior to the human mind in learning from past experiences through observed recorded outcomes, cause-effect analysis (causation), memory retention, memory expansion, and non-exhaustive compute power (abundance of energy). The human mind is capable of repurposing knowledge for evil. AI enabled computers are superior in their power to unleash enormous evil in the possession of tyrants and rogue states.
AI is a zero sum-game (balance), between the defenders (good) and the attackers (evil), and also an infinite game. Emotions differentiate humans from machines. Emotions are born from our inner fears, anticipated harmful outcomes, and basic survival instincts. Human emotions develop organically within years of birth. A fearless advanced machine has no emotions, only advanced algorithms. Machines, unlike humans, are not social creatures. Constitutional law and human rights do not (are difficult to) apply reciprocally to machines or software (data structures and algorithms). In Darwin’s theory of evolution, natural intelligence evolved from “natural” learning based on observed outcomes to actions (causation). In contrast, AI’s bottom-up evolution is based on a “pre-trained” learning model with programmable weights and biases.
Value creation with machine intelligence
Despite these challenges and risks, AI-powered intelligent machines can enhance the quality of life for everyone, provide livelihoods to wider population groups, and distribute digital power to all the people. Machine intelligence can help make things that regulate our activities of daily living more efficient, benefiting societies and the environment without displacing labor. Machine intelligence must illuminate human creativity, not eliminate human livelihoods. Applications where augmenting human-machine intelligence would be truly symbiotic are: smart transportation for fuel efficiencies, modernization of public infrastructure and utilities, environment protection and sustainability, precision instruments for surgery, advanced early diagnosis for preventive healthcare, sustainable agriculture for farmers, affordable housing for low income households, and overt surveillance for public safety and security.
Conclusion
As a global and interdependent society with a shared fabric of challenges and risks to overcome, technocrats and bureaucrats must re-think and strategize the path forward for positive use of artificial intelligence, and establish the metes and bounds to prevent unintended consequences of untamed technologies. This requires formalization of an etiquette for responsible use of AI. Defining the explicit objectives and methods for anticipated value creation that benefits societies, and evaluating the observed outcomes to mitigate negative consequences of the hypothesis, must become the first principle of AI.