The 'Morality' of AI Weaponry
February 24, 2021
Michael Brenner mulls the military admonitions of Silicon Valley's leading
lights.
A courtyard of Googleplex, the Mountain View, California, headquarters of
Google, 2014. (Runner1928, CC BY-SA 3.0, Wikimedia Commons)
By Michael Brenner
A congressionally appointed panel, called the National Security Commission
on Artificial Intelligence, is recommending that the United States military
launch a major program to develop autonomous weapons powered by artificial
intelligence (AI) software.
The panel, led by Google's former chief executive, Eric Schmidt, has
prepared a report that calls for the expenditure of $20-30 billion to
advance the program.
Its vice-chairman, Robert Work, a former deputy secretary of defense, said
autonomous weapons are expected to make fewer mistakes than humans do in
battle, leading to reduced casualties. "It is a moral imperative to at least
pursue this hypothesis," he said.
The draft report recommends that Washington "adopt AI to change the way we
defend America, deter adversaries, use intelligence to make sense of the
world, and fight and win wars." (Work is on the board of directors for
Raytheon, and board of advisers for Govini, a big data and analytics firm
committed to transforming the business of government through data science.)
Eric Schmidt, at right, in 2013. (Chatham House, CC BY 2.0, Wikimedia
Commons)
Categorical moral imperatives from On High is something I missed out on
while growing up. The cost of a misspent youth. How I envied our aging
Scottish neighbor who regaled me with tales from his childhood of attending
Calvinist services in Dunfermline. There, in church every Sunday, he was
instilled with the righteous obligation to improve himself by a steely
preacher of stern rectitude. I, by contrast to those youthful uplifting
spiritual experiences, whiled away the Lord's Day by lounging in front of
the TV watching ballgames.
So, I am thankful for the high-minded admonitions issued by Silicon Valley's
best and brightest that I have a "moral imperative" to support spending some
tens of billions on Artificial Intelligence directed weapons that kill only
those who deserve it, i.e. Washington's chosen "Elect." Now I have the
moral cause that I have been missing all these years - its absence being the
source of my irresolute vacillation on matters of violent death and
destruction.
Admittedly, some unfortunate habits of thought tenaciously resist being
wholly uprooted. The proclivity to test faith-based military doctrines
against reality is one of them. Awkward questions keep popping up to impede
my progress along the road of Ethical Truth and Redemption. Therefore, I am
earnestly beseeching your assistance in laying those doubts to rest. Here
they are, in summarized format.
The U.S. military's "shock and awe" bombing of Baghdad at the start of the
Iraq War, as broadcast on CNN.
No. 1 "Shock & Awe" The March 2003 bombing of hundreds of sites across Iraq
was intended to cow the Iraqi leadership into surrender as well as to clear
the way for the invasion. Large buildings, installations and complexes were
destroyed with thousands of casualties. Question: how would AI weapons
systems have affected this strategy? Is there a more precise version of
"Shock & Awe" which has the same psychological and military impact?
No. 2 Fallujah In the course of the war, American forces levelled most of
Fallujah - twice, driving a hundred thousand or so into the desert.
Question: given that the enemy were dispersed throughout the city, and
camouflaged, what would have been the capability of AI systems to identify
appropriate targets through algorithm-driven searches in data banks for
retina matches, facial recognition. etc? How might the civilians be alerted
by the guys at Fort Meade and at Central Command that there was no match for
them and that they safely could stay home watching CNN videos of the assault
on their TVs?
Nov. 15, 2004: Street in Fallujah heavily damaged by fighting. (U.S. Marine
Corps, James J. Vooris)
No. 3 Mosul & Raqqa In the campaign against ISIS, we razed most of Mosul and
Raqqa. The Marines alone fired 20,000-plus artillery shells into Mosul, a
densely populated city of well over a million. The Air Force flew
30,000-plus combat missions. The French air force, the Iraqi air force, the
National Army, and the Shi'a militia also fired on the city.
With all this firepower, it took several weeks to clear Mosul of ISIS
fighters. Civilian casualties and refugees have been too numerous to count.
Question: how would AI have enabled us to achieve the same objectives in
sanitary fashion in the light of ISIS tactics of evasion that were far more
complex than those encountered in Falluja?
(The other alternative for taking an urban area is fighting house-to-house
with only limited artillery and air support. That, of course, means higher
casualties for combat troops. Razing the city in effect substitutes civilian
casualties for military casualties among the attacking force - at a ratio of
10:1 or 20:1?)
An American soldier carries a wounded Iraqi child to a treatment facility in
March 2007. (James F. Cline III, Wikimedia Commons)
No. 4 Libya Here the U.S. provided the bulk of the firepower to topple and
kill the country's leader, Muammar Gaddafi. With his grip removed, the
country has succumbed to a decade of nearly continuous multi-party strife
with death, devastation and mass dislocation in its train. Currently, the
political scene is dominated by the power struggle between two warring
coalitions.
One, based in Tripoli, was installed by Washington and its allies, and has a
nominal status as the internationally recognized government. The other,
centered in the eastern city of Benghazi, is led by General Khalifa Haftar
who won favor by defeating some of the jihadist militias that have
flourished since Gaddafi's death. The former is backed by Saudi Arabia, the
UAE and especially Turkey's leader Recep Erdogan, who has transferred
thousands of his al-Qaeda, ISIS and other militant proxies from Syria.
Aerial view of Tripoli, Libya, in 2011. (MrPanyGoff, Wikimedia Commons)
Tripoli itself is dominated by Islamists. Haftar has the backing of
neighboring Egypt, Jordan, Russia and France. Various tribal militia and
jihadi movements roam the vast interior of Libya. The United States recently
has equivocated in its support while urging the two sides to reach a modus
vivendi - with little success. Question: Washington provides arms, directly
or indirectly, to ALL of the above-mentioned parties - except the Islamist
militias. Would its influence increase and its objectives have a better
chance of being realized were those arms deliveries to include AI weapon
systems?
No. 5 Yemen: The Saudi-led slaughter and starvation of Houthis has depended
crucially on the participation of the American military. Their campaign
would have been impossible without U.S. airplanes providing refueling, U.S.
officers sitting at Saudi command headquarters to direct targeting and U.S.
Intelligence. In addition, we provided the latest in "smart" bombs and
missiles. We also helped to enforce the blockade that denied food and
medicine to the general population. The Saudi intention has been to coerce
the Houthis into submission by causing as much misery as possible. Question:
how would extensive employment of AI weapons systems have changed any of
this? Would not the critical 'targets' be located in Washington, Tampa, and
Riyadh which are readily accessible to non-AI assisted weapons?
A destroyed Yemeni village. (Almogdad Mojalli/VOA)
No. 6 The Vietnam War This was a conflict between a country-wide
insurrection and a massive counter-insurgency campaign. A major part of the
American strategy aimed to deny the Vietcong sanctuary by denuding large
areas of vegetation (Agent Orange) and forcing civilians out of broad
swathes of territory declared as 'free-fire' zones. Would the possession of
AI weaponry have altered in a substantial way the conduct of the war,
casualty levels and the outcome? How?
No. 7 World War II The carpet bombing of cities became the norm. It
culminated in Hiroshima and Nagasaki. Civilian casualties in the earlier
fire-bombing of Tokyo were nearly as high. It was apiece with the air raids
against Hamburg and Dresden. They all shared the objectives of undermining
the enemy's morale, crippling industrial production - and satisfying the
lust for revenge against German air assaults on London, Coventry, Rotterdam
and Japanese atrocities. (As the invasion of Iraq was revenge for 9/11).
Question: had we access to AI weapons at the time, would this thinking have
been any different and actions otherwise?
U.S. aircraft on Tinian Island that took part in the Hiroshima bombing.
(Harold Agnew, Wikimedia Commons)
Reflection on these questions leads us to three clear conclusions:
No. 1 Widespread deployment of weapons directed by Artificial Intelligence
will have no appreciable change on decisions regarding the use of military
force, the conduct of war or on its consequences.
No. 2 Reviewing the history of warfare, or restricting our consideration, to
recent wars in the Middle East involving the United States, the fraction of
civilian casualties that might be reduced by the use of AI weapons is very
low. Those acts of war that produce the vast majority of casualties are not
of a kind that allow for the large-scale substitution of AI arms. *
(AI directed weapons will be developed, deployed and used at an accelerating
rate - nonetheless. In our post-modern age, any high-tech innovation
associated with electronic gadgetry enjoys a presumptive claim to superior
utility, value and virtue.)
No. 3 The innovation associated with AI arms, like drones and the volunteer
army, could encourage military activism by promising fewer American
casualties. The net result, though, could well be more net death and
destruction in aggregate.
Several years ago, I attended the Republic Day ceremonies in New Delhi. I
had a curbside seat for the parade led by Nehru and his cabinet along the
grand Raj Path from the Presidential Palace (Rashtrapati Bhavan) to India
Gate. They were followed by other dignitaries, floats, and then an
impressive array of turbaned cavalry in magnificent attire. A spectacle
worthy of an Imperial Durbar. Indeed, Lutyens had designed it all to honor
the Immortal British Raj.
At the very end of the procession, trailing the horsemen at a respectable
distance, was a less glamorous squad of khaki-clad "sweepers: who used straw
brooms to clean up the manure left behind.
Today's political world brings back memories of that occasion insofar as we
are constantly presented with a huge litter of BS in the aftermath of the
latest mega event, and lesser amounts after smaller events. The urgent
recommendations in the AI report by the commission falls into the latter
category, but lack nothing in pungency.
Overall, there is an imbalance of production and disposal. Shoveling
capacity hasn't kept up. The dilemma for commentators is being placed in the
position of New Delhi's "untouchable" sweepers who are required to do the
dirty work of keeping the public space clear of odiferous matter.
Repetition makes the task all the more irksome. It's as if the 800-strong
Light Brigade, commanded by our home-grown Lord Cardigans, passes through
daily on march to the Valley of Death. Yes - somebody has to do it,
"untouchables" don't exist, and the best solution would be to stable the
"horses." The chances of the last happening any time in the foreseeable
future are nil.
I personally have taken leave. So, instead of a systematic dismantling of
the report, I have followed the avoidance strategy of restricting myself to
the few remarks and questions presented above.
* The pitchmen for AI weaponry base their promotion on a mistaken conception
of warfare. They presume that it is essentially instrumental in nature. It
is not. Destruction, looting and brutalizing civilians is an integral part
of war - usually.
At the extreme, think of Attila the Hun whose armies slaughtered and
pillaged routinely, or Genghis Khan's Golden Horde who did the same, or
Timur whose calling card was a pyramid of skulls at the city gate, or the
Nazis. For them, antiseptic warfare would have been anathema - taking all
the fun out of it.
The passions of wars inspired by religion or other ideologies also bode ill
for non-combatants: e.g. Joshua's annihilation of Jericho - obedient to
orders from his Commander; the Crusades; the 17thCentury's Thirty Years War;
the Islamic State's recent rampage. Or the scourge of modern nationalism.
Were the Serbs shelling Sarajevo concerned about precision targeting? Would
Radovan Karadzic or Slobodan Mladic have welcomed AI weapons as a solution
to a moral dilemma that endangered their immortal souls?
In all of these instances, the very notion of "collateral damage" was
irrelevant.
Michael Brenner is a professor of international affairs at the University of
Pittsburgh.
