Smart Artificial Intelligence

Alan Turing, computer genius of World War II Britain, posited in a 1950 paper that machines could possibly become as intelligent as humans.  The so-called Turing test put forward a scenario where development of computers would come to a point where a neutral observer could not tell the difference between a human and a machine in blind test responses. A machine could actually “think.”

It’s been controversial proposition since day one, but it is a foundation principle in the 21st century development of Artificial Intelligence (AI). And for the U.S. military, private sector businesses, and our political leadership, AI is today strongly enticing, quite opaque, and a potentially dangerous matter sitting right on our doorstep.

The Realities and the Concerns

AI today is nowhere near the Turing Test.  But, it is rapidly moving there with speed enhance by the promise of quantum computers which move the “thinking” software and algorithms on which AI is based far faster than any human could respond.  In an age of developing hypersonic weapons, ever more complex consumer supply chains, threatened and broken software encryption, and instaneously moving fake news, even these early versions of AI could provide a real tool to counter these problems and more.

The recent U.S. Commission on Artificial Intelligence, chaired by former Deputy Secretary of Defense Bob Work and Google chieftain Eric Schmidt examined the challenges of AI and the use of it by other nations around the world. The report painted a complex picture of the need for American development of AI.

The Commission noted our vulnerabilities in use of AI in warfare,  Intelligence gathering and analysis, and finding the right kind of talent to enable the advancement and use of AI.  It also expressed concern over the leakage of our technology to other nations (read: China – a nation that considers AI a top priority) and the challenges of our microelectronics industry staying ahead of the game. Oh yes, and a loose Intellectual Property system that does not protect American inventions from overseas poaching.  Add to that the age-old problem of needed additional investment in the development of AI from the government and the U.S. has some real AI headaches.

Person in the Loop and Bias

AI represents several other less tangible, but still important challenges and concerns.  For the military dealing with increasingly fast and complex weapons, AI represents a real countermeasure to such things as hypersonic missiles.  And, of course, we can use AI for our weapons.  The problem, however, is what will these weapons do undirected.  Humans still hold the lead in complex judgment.  How do you keep the human in the process and not retard the system’s effectiveness?  

Autonomous weapons, for instance, represent a real ethical problem.  A human in the loop would be able to make the judgement about when to fire a missile on a target based on whether the target could be hit without hitting other humans. What is the “value” of potential “collateral damage” – for instance, are they children, or family members, or simply in the wrong place?

Another big problem is bias in the AI systems themselves.  AI is based on software created by human beings who develop the algorithms for the use within the software. In short, it carries within it all the biases of the humans who wrote it.  As of now, the industry is dominated by white, middle class, “educated” males in their 20’s and 30’s. Does this lead to faulty choices when used such as predicting potential high crime areas to readjust police resources and focusing on ethnic neighborhoods? Indications are that it does.  Equally facial recognition seems to be a problem based on these “baked in” biases.

What Now for AI?

The AI Commission Report is a good first step in identifying problems and potential solutions.  However, the ground in D.C. is littered with bright reports done by bright people.  The best hope is that the tech-minded Biden Administration will go beyond picking up only the rhetorical charge on AI and make it an important part of our national security strategy – as currently does Beijing. 

We need to set rules for its use at home and among like-minded nations.  And we need to support its development in the U.S. --  providing R&D support, training a new generation of AI experts, and substantially improving our legal protection of our IP.   

Welcome to the 3rd decade of the 21st century.  Technology matters more than ever in national security policy.  And AI is the way of the future.

Ronald Marks is Term Visiting Professor, George Mason University, Schar School of Policy and Government. He is President of ZPN Cyber & National Security Strategies                                           

Image: Unsplash

You Might Also Read:

A Successful Solar Winds Investigation: