Ethics in Machine Learning

October 30, 2017

Rise of the Machines

Sentient AI overlords are a ways off
There are real concerns we need to be thinking of right now

Unintended Consequences

In the Terminator franchise, the machines weren't frightening because they were learning from data and acting on it. (That is the purpose of AI, afterall.) They were frightening because they were learning from data and acting on it in an unintended way. These "unintended consequences" are one of the biggest ethical dilemmas facing the AI community today. Of course, this can happen because of a programming error, as depicted here.

(And if you don't see the problem here, it's because we've used an assignment operator instead of an equality operator, which effectively tells the robots to kill humans.)

More likely, with AI, the error will occur either while specifying or training the model.

Enter: Microsoft Tay.

Reference: https://www.reddit.com/r/funny/comments/4x28gw/this_is_why_code_reviews_are_a_good_thing/?st=j770kl8d&sh=9fb83792

Unintended Consequences

Unintended Consequences

Now, you might wonder why Microsoft would unleash a bot upon the world that was so unhinged. Well it looks like the company just underestimated how unpleasant many people are on social media.

– Ashley Rodriguez, Quartz

Model training error
Data quality issue
Narrow focus on technology while neglecting actual users

Biased Data

Biased Data

Lack of diversity in training data
Possibly unintentional
Engineers must be aware of implicit bias

Safety Concerns

AI can improve safety!
The more we trust, the better our AI has to be
Users must be aware of the risks

Safety is another major concern. It should be noted that AI can greatly improve our safety! From Tesla's crash prediction and warning system, to the federally mandated Positive Train Control program designed to stop head-on collisions between trains, AI can do things we as humans cannot, and it can step in when humans are prone to error.

It must be understood though, that the more we lend our trust to AI, the more we need to be aware of the inherent risk. Something like a Roomba may pose no greater harm than scaring the cat, but as we move up to something like Tesla's autopilot, the stakes increase dramatically. One man has already been killed because he was watching a movie while trusting Tesla's autopilot entirely. Though the system warned him as many as 7 times, he ignored the warnings and was killed in a crash. Not only must we, as engineers, integrate safety features into our algorithms, but we must make sure our users know the risks and how to operate our products safely.

References: https://www.cnbc.com/2017/06/20/man-killed-in-tesla-autopilot-crash-got-numerous-warnings-report.html

Intentional Harm

Asimov's Three Laws of Robotics:

A robot may not injure a human being or, through inaction, allow a human being to come to harm.
A robot must obey orders given it by human beings except where such orders would conflict with the First Law.
A robot must protect its own existence as long as such protection does not conflict with the First or Second Law.

You may be familiar with Isaac Asimov's Three Laws of Robotics. Asimov developed the laws in the 40's and they have been the subject of much philosophical debate. But the fact of the matter is, these are not actually laws enacted by a government or international community yet, but rather guidelines by a science fiction writer. In reality, autonomous robots that cause harm are already here. We certainly have the technology to develop them. There are extensive debates to be had about the laws we put around autonomous machines that are made to harm humans. The topic is largely unprecedented, perhaps closely rivaled by something as outlandish (but not inconceivable) as space war. The only guidance we will have going into this uncharted territory is our existing grasp of ethics. If we haven't developed an ethical foundation yet, we are going in blind.

References: https://www.cbsnews.com/news/60-minutes-autonomous-drones-set-to-revolutionize-military-technology/ http://www.popsci.com/international-community-is-about-to-debate-killer-robots http://www.popsci.com/ban-autonomous-killer-robots-inevitable http://www.popsci.com/autonomous-weapons-are-already-here http://www.vocativ.com/430867/artificial-intelligence-algorithmic-warfare-drone-strikes/

The Need for Ethics

Through some of the previous case studies, you can begin to understand that there is a need for ethics in some form. Unfortunately, this is where it gets tricky. We begin to part ways with the binary ones and zeros of machines and stumble into the nebulous universe of philosophy. Aristotle argued in the first book of his Nicomachean Ethics that we can all agree that the goals of our actions is to attain some "Good". But our agreement on what is "Good" is where we differ.

My goal here is not to engage in philosophical debates on ethics, as that is not my area of expertise, nor is it a solved problem, even thousands of years since that question was first asked. My goal here is to argue that we need to begin thinking about it now. With the advancements in processor technology, the sophistication of machine learning algorithms has improved drastically in recent years. While it feels like we have had AI around all along, we are, in reality, at the forefront of truly useful AI.

References: https://hellenicnews.com/aristotle-philosopher-invented-science/ https://hbr.org/2016/11/teaching-an-algorithm-to-understand-right-and-wrong Why Ethics Matters: A Defense of Ethics in Business Organizations, Velasquez, Manuel, Business Ethics Quarterly, Vol. 6, No. 2 (Apr., 1996), pp. 201-222

Regulating AI is Hard

I keep sounding the alarm bell, but until people see robots going down the street killing people, they don’t know how to react, because it seems so ethereal.

Elon Musk

This quote from Elon Musk should capture everyone's attention. Musk's endeavours are using some of the most advanced AI out there, so he is as credible source as one could be on the subject.

David Ha, a researcher on Google Brain, is no less credible, and he believes the

Governments will have a hard time regulating AI, though. In October of 2012, an Italian court sentenced six scientists to a six year prison sentence for manslaughter for failing to predict an earthquake that killed more than 300 people. Scientists argue that this sets a dangerous precedent that may very well discourage other scientists from important work for fear of retaliation for wrong answers. There is a balance to be sought, of discourage negligent behavior but encouraging research and discovery. This is likely to be an uphill moral battle.

https://www.theverge.com/2017/7/17/15980954/elon-musk-ai-regulation-existential-threat https://twitter.com/hardmaru/status/886673097981476865 Sisto, Alberto. Italian scientists convicted over earthquake warning. Reuters. https://www.reuters.com/article/us-italy-earthquake-court/italian-scientists-convicted-over-earthquake-warning-idUSBRE89L13V20121022 http://www.sciencemag.org/news/2015/02/why-italian-earthquake-scientists-were-exonerated

Not a Solution, but a Start

Computer scientists and engineers must examine the possibilities for machine ethics because, knowingly or not, they've already engaged in some form of it.

The Nature, Importance, and Difficulty of Machine Ethics, J.H. Moor

I don't claim to have an answer to the problems we face, but I do believe that integrating ethics training into the core curriculum of information sciences can have a huge impact. The UNO Computer Science and MIS programs require a course in IT Ethics, but unfortunately, the Mathematics program does not. Taking this small step could have a big impact on future generations of researchers and engineers.

Above all, the pioneers in Silicon Valley who are forging the bleeding edge of AI need to take it upon themselves to institute strong ethical standards in their organizations. Integrating ethics standards into performance reviews and creating an overall culture of high ethical standards sets a solid foundation on which AI can thrive without causing harm, be it intentional or unintentional.

But the time for consideration is now. If we wait any longer to bring ethics into the conversation, it may soon be too late.

References: Moor, J.H., The Nature, Importance, and Difficulty of Machine Ethics, IEEE Intelligent Systems, Volume 21, Issue 4 July-August 2006 http://ieeexplore.ieee.org/abstract/document/1667948/

Rise of the Machines

Unintended Consequences

Unintended Consequences

Unintended Consequences

Biased Data

Biased Data

Safety Concerns

Intentional Harm

The Need for Ethics

Regulating AI is Hard

Not a Solution, but a Start

Thanks!