by Greg Simpson

The societal impact of AI

May 16, 2018
Artificial IntelligenceEmerging TechnologyTechnology Industry

If we are going to augment humanity with the machine, we need to do it in a way that doesn’t bring along our mistakes of the past.

In “Merging with the machine,” I highlighted how the exponential acceleration of technology, combined with a focus on augmenting humanity with technology is driving a future where we begin to merge with the machine.

In fact, as I pointed out, we are already augmented, by the smartphone you may be reading this on. This can drive significant changes to our time on earth, both positive AND negative.

Technologists are often optimists by nature; it is a necessary trait to enable innovation. Otherwise, the “fail fast” model of innovation would quickly result in the innovator giving up and taking up a new career! The danger is that if all of our technologists are optimists, then who is thinking about the negative impact of the introduction of technology?

Given our current news cycle, some of these potential negative impacts are now clearer. (Think Zuckerberg at the Senate.) This blog is my attempt to layout some of the bigger pros and cons of our rapid push toward an augmented society.

In my view, there are four large classes of pros and cons. Let’s consider the list.

1. Systematic decisioning/bias

Well-designed systems can help us remove bias and promote diversity in our world. Examples include resume screening tools that remove names and prevent unconscious bias from factoring into candidate pools. The danger is that our current data pools and our current technology already has a great deal of biased influence. Our male dominated technology field develops most of these tools.

At the Neural Information Processing Systems conference, 90% of the papers were written by men. If we use biased data to train our machine learning AI, then we get biased AI. The classical garbage in, garbage out problem. Often the bias is unconscious or is based on societal norms that have formed due to preconceived notions that have historical roots. Simple examples are displayed in our virtual assistants. Most use a female voice by default. This furthers the gender biased stereotype that assistants are often women. We need to actively work to ensure that the data we train our bots on, and the decisions we make about how we deploy voice interfaces and even the names of the bots not continue to propagate gender stereotypes.

2. Free access to info/filter bubbles

The internet was thought to make the world flat. Thomas Friedman’s famous book The World is Flat talked about how the advent of global communications, like the web, enabled the delivery of intellectual property from anywhere on the globe. However, with the advent of social media algorithms that feed you information that is tuned to what you “like” … more and more people are living in “filter bubbles” that reinforce their own biases. Even reinforcing things that are factually untrue. The world is not actually flat. It is a sphere. This is a well-documented fact.

However, on the internet, if you think the world is literally flat, you can suddenly find a thousand other people (out of the billions on-line) that agree with you. These algorithms can then bring you more and more data to help you reinforce these false beliefs, simply because you like to read about this new found “fact”. Hence the world is flat digitally… but not literally. Free access to information is a benefit, but also a curse when it allows a falsehood to gain equal footing to the truth.

3. Equality/inequality

There is hope that technology can afford more people opportunities, and we see that in some of the startups that are created around the world. For example, technology that enables the rural African farmer to understand market pricing for their goods so as to not be taken advantage of by a larger, more informed buyer. Sheryl Sandberg, COO of Facebook said, “If we can extend [the internet] to more people, we increase voice… we increase economic opportunity… and we increase equality.” This is the hopeful view.

At the same time, there is concern that economic inequality could grow even more and take away the ability for one social economic class to benefit from technology. The Financial Times said, “Silicon Valley fears vilification … for prospering at expense of everyman.” Will Silicon Valley’s riches only come to the few, or will technology enable a flatter, more equal world?

4. Social connections/social isolation

We’ve all seen the family/couple/friend group at the restaurant who has lost the art of conversation. Instead, they are all gazing at their phones. Silently communicating by typing to a completely different disconnected group of friends. Social media helps us connect with old friends who may live in distant locales, enables the grandparents to see a constant stream of pictures of their new grandchild who is located thousands of miles away, but it can also drive loneliness and a loss of social graces. I, for one, do not want to make Gartner’s prediction, that by 2020, people will converse more with their digital assistant than with their spouse, true.

In summary, if our emerging technology is going to merge humanity with the machine, augmenting our human intelligence with artificial intelligence, we need to manage the risk! The risk of isolation, the risk of losing our art of conversation, the risk of convincing ourselves that our pre-conceived biases are not only factually correct, but good.

The value that technology brings to the world is worth the risk. I am, after all, an optimistic technologist. But we need to make sure we are bringing diverse skillsets and viewpoints into our development of this new frontier. Otherwise, in a world that only accelerates every day, we could accelerate into a dystopian future (as illustrated by numerous books and shows… just watch a few episodes of Black Mirror) instead of into a future where an augmented humanity is a better world for all, not just the few, or for the machines.