The term Artificial Intelligence (AI) is generally misunderstood. So a clear definition would help. “Artificial intelligence (AI) is the ability of computer systems to perform tasks that normally require human intelligence—such as learning, reasoning, problem‑solving, perception, and language understanding—using data‑driven algorithms and computational models.”
Indeed, this definition was derived from—you guessed it—an AI program developed by Microsoft called “Copilot.” Even though these programs are no substitute for human judgment, they’re useful for aggregating internet data.
These assistants do not represent the full power of AI systems or what they can potentially do to assist mankind. Indeed, they represent hardly a bite-sized hors d’oeuvre compared to the plat de resistance that will be served up as AI systems are perfected in the future.
AI can advance medical research and cure diseases that have haunted mankind for millennia. Future generations can also benefit from AI in education, transforming learning from cramming students together in the classroom to individually tailored and accelerated learning opportunities for those aspiring to a profession that suits their goals.
AI’s role in streamlining industry will also spark a productivity revolution that will wildly outpace the Industrial Revolution of the late 18th and 19th centuries when machinery, steam power, the growth of factories, and the mass production of manufactured goods swept the Western world.
A less noble aspect of the AI revolution concerns an area of research that has occupied me for the past three years: a revolution in AI-enabled warfare. AI in military operations will give commanders in the field a distinct advantage over their opponents. It will enable commanders to act on exquisite information with remarkable speed before opponents realize they are being targeted. Constellations of sensors will make the battlefield more transparent than ever before. And AI will be essentially in the business of sorting volumes of targeting information that, when attacked in the right sequence with the best munitions, will deliver a certain victory.
Unfortunately, there has been precious little useful discussion now that AI has become the latest political football in America. The AI discussion today centers on how this superintelligence will usher in the end of humanity. In a word, it’s hogwash. Yet there are genuine concerns about how to ensure AI doesn’t develop to the point where humanity cannot control it. For that reason, we need more people like Dr. Elise Annett and Dr. James Giordano to inform the discussion.
These two ethicists have given a lot of thought to the AI debate. Americans are being ill-served by hysterical politicians and, sadly, self-serving AI CEOs and techno-poohbahs. The former seek political victories at the polls. The latter feign AI disaster to entice government involvement in AI production so they can skirt product liability by blaming the government for letting them develop programs that wind up hurting people. In other words, the debate we are hearing is both opportunistic and deceitful. It needs to be elevated by logic and critical thinking.
In their article titled “Between Extinction and Hubris: A Dialectical Approach to AI Risk, Control, and Governance,” which appears in the Institute for National Strategic Studies, Annett and Giordano make this observation.
“AI remains extraordinarily capable but importantly bounded. It does not possess independent political authority, military command, economic sovereignty, or generalized physical agency.” They recognize that AI has vulnerabilities, including others’ ability to manipulate it and the possibility of unpredictable behavior. But their larger concern is the risk that humans will progressively cede to AI because it will be “efficient, economically advantageous, operationally expedient,” and, in a military context, viewed as essential to outthinking the enemy. In other words, AI will be so good that humans will let it run wild. How do they suggest we govern AI?
They wisely assert that “[A]ny posture toward effective governance should address AI capability, human cognition, organizational behavior, institutional incentives, and geopolitical competition as components of a single evolving system; because the essential question is whether humans and their institutions will retain the knowledge, discipline, humility, and capacity required to exercise such control when it matters most.” Bingo.
Among their many recommendations are (1) regulations that address what a system can actually accomplish rather than the title it’s given; (2) mandate ongoing testing and evaluation to ensure AI doesn’t create harm; (3) require that humans possess adequate information, competence, time, and actual authority to challenge or abort AI-generated actions; (4) ensure that as machines become more capable, requirements are adjusted to control them; (5) create chains of human responsibility; (6) require mandatory AI incident-reporting systems; (7) create independent high-risk AI oversight mechanisms; and (8) establish international norms for the employment of military AI.
Indeed, what we need in America is less hysteria about AI development and more sound, authentic logic from critical thinkers like Doctors Annett and Giordano to frame it.
0 Comments