By Amirteymour
Moazami (Spring 2018)
How Artificial intelligence (AI) & Machine
learning (ML) are disrupting traditional industries but yet can pose a threat
to humanity
Imagine a
world where cars can drive themselves, reduce traffic, and increase safety. Imagine
a world where robots fight wars instead of soldiers. Is AI dangerous or
convenient? Historically, ideas like this were just part of one’s imagination,
yet, today, AI & ML are making these once science-fiction ideas become a reality!
What
is the difference between AI & ML?
Artificial
intelligence is software that performs tasks which involve intuition and would
usually require human intelligence. Human intelligence: our ability to reason,
make decisions, form plans, understand information, which can all be done with
AI. In other words, AI helps computer systems to think like humans. Do you know
Siri, the smart personal assistant on an iPhone that helps you get things done
by using speech recognition? (Gruber, 2017) That’s the power of AI.
Artificial
intelligence gave birth to machine learning. The aim of ML is to teach
computers how to learn and act on their own by feeding it information and
finding patterns in the data. ML helps computerized systems get smarter over
time through experiences without depending on programming —
a technique that is known as unsupervised learning. Imagine you are teaching an
airplane how to fly. Every time the plane successfully takes off and lands you
say, “good airplane” and every time the airplane crashes you say, “bad
airplane.” This example of machine reinforcement learning is dependent on rewards.
Every time the airplane performs the correct task it gets a positive reward.
Over time, the machine understands how to work itself to maximize positive
rewards (Ng, “Machine Learning Lecture”). Another powerful example is how Tesla
learns over time as they collect more and more data from cars driving on
streets while in assisted-driving mode. As machines become more knowledgeable,
so do humans.
Is
AI a blessing?
AI & Driverless cars
Source: Qualcomm & TomTom
Self-driving cars
are here. Waymo, Google’s driverless
car division, has been laser-focused on developing its self-driving technology
since 2009. They are committed to creating a safer driver that is fully aware
of its surroundings at all times and never gets distracted, so people and
things can move around effortlessly without the need for a human driver. Waymo is
working hard to offer a reliable and cheap source of transportation to
everyone, making roads safer, while still being eco-friendly. Google’s self-driving
vehicles have already obtained more than 4 million miles of driving experience
on city streets in areas like Austin, TX and Phoenix, AZ. This significant
amount of self-driving experience from Google’s fleet of vehicles implies that
these cars are continuously learning on their own by collecting information.
And this data helps the driverless car control all sorts of real-world
situations and conditions in a safe and non-disruptive manner. AI & ML is
an important part that has enabled Waymo to make significant progress on
driverless vehicles, which intrigues me. Intriguing is: ability of computers to
learn, behave, and render judgment almost as good as humans.
How
self-driving cars work?
Self-driving vehicles
have sensors, radars, cameras, and software within them that is programmed to detect
road work, other cars, cyclists and pedestrians up two hundred yards away in
all directions. Given that self-driving vehicles are computerized systems, they
are continuously gathering information that updates to the cloud in real-time,
which allows the car to become smarter about its environment the more miles it drives
(“Waymo Safety Report on the Road to Fully Self-Driving”
12-15).
For example, if
your self-driving car’s sensor sees that another vehicle in front of it is
slowing down, then the software predicts that the car in front of you is
approaching a stop sign or a red light. This observation would cause your car to
stop comfortably to avoid an accident. Or imagine crossing the road while you are using your phone. A car is
driving towards you, and you’re unaware of it. But all of the sudden the
driverless car detects you and slows down automatically. Wouldn’t you and your
family appreciate that?
Safer
streets
According to the
Waymo Safety Report, “94% of crashes involve human choice or error” that led to
1.25 million deaths globally in 2014. Accidents caused by humans are usually
because the driver is either drunk or distracted. So, developing driverless
cars that are always paying attention, learning in real-time and making roads
safer can cut down accidents by about 90% and save $190 billion in healthcare
costs as told by a report conducted by McKinsey (Ramsey 2015). And this is all
possible because of AI and ML. Isn’t this awesome?
Or
is AI a threat to humanity?
Will intelligent
machines control humans, or will we control intelligent machines? AI systems can
become smarter than any human and outperform them in almost all intellectual
domains, which means that it will be nearly impossible for humans to predict
how these systems will behave. As machines learn to think for themselves and
adapt to their immediate environments, they present a substantial risk to
humanity.
Autonomous
weapons
Weaponized
AI is built to cause disruption. Autonomous weapons have artificial
intelligence systems within them and can be in the form of guns, robots or
drones. Intelligent weapons have the capability of carefully choosing and eliminating
targets (people and establishments) without human intervention. If autonomous
weapons are in the hands of the wrong people, it can result in a large number
of deaths or injuries without difficulty (“Benefits & Risks of Artificial
Intelligence). For example, an armed drone is a plane that does not have a
human pilot and can damage targets based on a specific pre-established criterion.
Militaries
around the world realize the potential of AI and how it can be a useful tool in
warfare. If any dominant military organization around the globe follows through
with the development of autonomous weapons, an AI arms race is bound to happen.
According to Russian President Vladimir Putin, “the leader in AI will rule the
world.” An AI arms race between militaries can lead to a high death rate
globally. An AI war is not helpful for human advancement. But using AI to make
the combat zone a safer place for people, especially for ordinary citizens is
beneficial to humanity. According to Dr. Peter J. Bentley, a computer scientist,
and professor at University College London, “We can reduce casualties in
conflicts since losing a drone is better than losing a person” (Allison, 2017).
He says it is easier to use drones in warfare rather than a person since the
cost of human life will never equal the value of weaponized AI regardless of
how pricey it is.
How
to manage a future with AI?
AI can pose a risk
to humanity. But it can also help solve diseases such as cancer using genomic
data and let robots fight wars instead of humans. Being mindful of the dangers
presented by AI and applying best practices possible to manage its potential
consequences through proactive regulation can make AI beneficial for humanity. It’s
our decisions as humans to decide whether AI will compete with us or help
satisfy our needs and improve our well-being.
References
Allison, Peter Ray. “Will Killer Robots Make Us Safer?” Alphr, 14 Dec. 2017
www.alphr.com/arificial-intelligence/1007974/will-killer-robots-make-us-safer
“Benefits & Risks of Artificial Intelligence.” Future of Life Institute, futureoflife.org/background/benefits-risks-of-artificial-intelligence/.
“How AI
Can Enhance Our Memory, Work and Social Lives.” Performance by Tom Gruber, TED, Apr. 2017, www.ted.com/talks/tom_gruber_how_ai_can_enhance_our_memory_work_and_social_li ves.
Karpukhin, Sergei. “Putin Says the Country Who Leads the
Development of A.I. Will Rule the World.”
CNBC, 4 Sept. 2017, www.cnbc.com/2017/09/04/putin-leader-in-artificial-intelligence-will-rule-world.html.
Ng, Andrew. “Machine Learning Lecture.” Palo
Alto, California, Stanford University.
Ramsey, Mike. “Self-Driving Cars Could Cut Down on
Accidents, Study Says.” Wall Street Journal, 2 March 2015, https://www.wsj.com/articles/self-driving-cars-could-cut-down- on-accidents-study-says-1425567905
Walsh, Toby. “AUTONOMOUS WEAPONS: AN OPEN LETTER FROM AI
& ROBOTICS RESEARCHERS.” Future of Life Institute, 28 July 2015, futureoflife.org/open-letter- autonomous-weapons/.
“Waymo
Safety Report on the Road to Fully Self-Driving.” Waymo, pp. 1–42., storage.googleapis.com/sdc-prod/v1/safety-report/waymo-safety-report-2017.pdf.
No comments:
Post a Comment