The car factory that employed a woman whose skull was crushed by a
robot that encroached on her workspace blames the machine. The jury, as they say, is still out.
Equifax Inc. (NYSE: EFX) blames its
software for exposing the highly sensitive details of the credit company’s 143 million users. The software firm fires back and says
it’s
a people problem.
Facebook Inc (NASDAQ: FB) blames an
ad-targeted demographic meant to appeal to anti-Semitic users on an out-of-control algorithm. Not the
human who designed it.
Tesla Inc (NASDAQ: TSLA) blames a
dead driver for a crash,
someone who was haplessly seduced into believing the hype of the company’s semi-autonomous smart vehicle. The feds blame Tesla.
Related Link: How
The US, China And Russia Are Weaponizing AI
Does that make CEO Elon Musk — the world’s leading alarmist on the threat of
artificial intelligence autonomy who pressures his people to make deadlines
on his rockets and self-driving cars — the ultimate
hypocrite of the smart tech era?
You be the judge. Everybody else is. We are entering an era of machine autonomy, and even the lawsuit
industry is choosing sides and sizing up how to assess liability when the Internet of Things goes horribly wrong.
Blaming Tech Is The Big New Problem Of The AI Age
As tech gets smarter, the race to unleash AI presents new opportunities for hackers to leave untraceable tracks in new
algorithms created by scientists driven to get autonomous systems to market.
And when machines designed to teach themselves make bad decisions, the reasons why often elude human grasp, leading people to
blame their creations instead of their creators, says Eugene H. Spafford,
Purdue University professor of computer science, philosophy, and founder of the Center for
Education and Research in Information Assurance and Security.
The Problem Of Tracking AI Errors
“It’s an area of big concern. Not just cars but UAV (unmanned autonomous vehicles), trucks. We should be concerned about how
those things are put together,” Spafford told Benzinga. “If one of these things results in death or injury, who would be
responsible? I’m not sure some of the organizations pushing the technology have given it sufficient thought.”
One of the problems with emerging systems, of machine learning in general, is tracing how they came to the decisions that often
surprise the scientists that feed them massive amounts of data.
Spafford suggests that autonomous cars — already proven to be eminently hackable — might even elude an expert investigation of
whether robotic behavior was even a result of a hack.
Some Scientists Try To Head Skynet Off At The Pass
While some scientists race toward the autonomous world of AI and robotics, some are trying to keep up with the safeguards
necessary to keep, say, massive hacking of car fleets from matching the enormous numbers of highway deaths caused by human error, a
Darwinian trade-off if ever there was one.
The Defense Advanced Research Project Agency has created the Supply Chain Hardware Integrity for Electronics Defense program to
enhance supply chain traceability.
Northrop Grumman Corporation (NYSE: NOC) is
leading a team to develop tiny chips to enable traceability. Spafford said DARPA is delving deeply into the whole realm of
traceability, figuring out why autonomous systems sometimes make surprising decisions.
“One of the big problems is being able to go back and find out why did they do the thing it did,” Spafford said. “Even why it
did something legitimately. Why did it treat the stop sign that way? It’s difficult to go back and trace that.”
The University of Washington recently did a fairly frightening experiment. It put pieces of tape
on a stop sign, and convinced a self-driving car it was a speed sign. It was a low-tech way to seriously mess with the head of
state-of-the-art autonomy.
“That suggests that if someone finds a way to hack in and do it, we may not be able to find out why it happened or who did it,”
Spafford said. “I’m concerned about these things being done accurately.”
Holding Algorithms Accountable
Leave it to the legal community to already start divvying up how to work the liability angle. Right now, as usual, the law lags
behind tectonic jumps in tech.
“Traditional tort law would say that the developer is not liable,” wrote Jeremy Elman, a partner at the global corporate
legal giant DLA Piper and executive of the firm’s Intellectual Property and Technology and Emerging Growth practices. “That
certainly will pose Terminator-like dangers if AI keeps proliferating with no responsibility. The law will need to adapt to this
technological change in the near future.”
For its part, the computer scientists
are sending warning signals.
To virtually no coverage by the media at all, the preeminent Association for Computer Machinery on Sept. 14 issued a demand for
algorithmic accountability, in essence a warning that humans are
hiding their blame for things gone wrong behind the machine-learning technology they are creating.
So while Elon Musk harps endlessly on about the threat of AI and warns that humans must incorporate cyborg components to
compete, or flee to Mars, he is a partner in making the world of the auto-machine future.
And while the moral philosophers wonder about the dangers of AI, the militarized nations are racing to weaponize AI for the next
generation of killing
machines.
“There are people who don’t care about the dangers,” Spafford said.
Related Link: Why A Charming
Synthezoid Is Coming Your Way
© 2017 Benzinga.com. Benzinga does not provide investment advice. All rights reserved.