How do you eat an elephant? One bite at a time...
That’s exactly what Mustafa Suleyman and the team at DeepMind did this week when they took the first step to address the proverbial elephant in the room: the potential ethical and societal implications of AI.
DeepMind just announced the formation of a new research group called DeepMind Ethics & Society (DMES). This group is made up of dedicated staffers and a number of unpaid external “fellows” including Nick Bostrom, Coyle, and Jeffrey Sachs. Their focus is to tackle the biggest issues in artificial intelligence. This includes any problems with managing AI bias, the coming economic impact of automation, and the need to ensure that any intelligent systems we develop share our ethical and moral values.
This is coming at an important time where the AI community is receiving a lot of blowback given Elon Musk’s recent suggestion that AI is the biggest risk we face as a civilization. While I think much of this is misconstrued, it doesn’t stop the constant barrage of sensational headlines adding to the fear mongering.
The good news, DeepMind has shifted the focus from conjecture to pragmatism. Right now, the single biggest roadblock in advancing AI has less to do with superintelligence and more to do with the human factor.
Let me elaborate a bit and illustrate with a few examples:
If we are diagnosing disease, we need to be extremely careful not to assume an inherent bias based on a limited cohort, e.g. the US population. This would ignore the rest of the world, thus causing it to be irrelevant or result in greater harm. Instead, we must surrender to the practice, just as you might do in a meditation. In other words, we need to be extremely mindful as we are training new algorithms to avoid consciously, or unconsciously interfering with a preconceived bias.
Fully autonomous cars are expected to hit the market within the next 5-10 years. This innovation will likely disrupt multiple industries, and force layoffs particularly with taxi and truck drivers. It will be important to pass legislation to not only to address unemployment, but to also encourage and reward innovation. While this seems like an inevitable next step, many legislators in Washington choose to ignore this reality and are too consumed by partisan politics to make any progress.
The top 10% of families hold 76% of the total wealth of the U.S. Unfortunately, this statistic isn’t moving in the right direction. We need to democratize AI to level the playing field, so that no country (or individual) is caught behind the eight ball. Of course, this means we need find creative ways to make AI more cost-effective and accessible. Consider the smartphone evolution: according to Pew Research, about three-quarters of U.S. adults (77%) say they own a smartphone. This is up from 35% in 2011 because smartphones have become cheaper and more intuitive.
As you can see these points merely scratch the surface as we uncover the real and immediate issues associated with AI. Clearly, we have much bigger fish to fry before we concern ourselves with a cybernetic organism like the Terminator! This shift in thinking will require inclusive leadership, good intentions, and the desire to explore all possibilities. I think we need to embrace what DeepMind is doing and find a role to contribute and advance the cause. As Suleyman recently said, “We want these systems in production to be our highest collective selves.”
I can only imagine we will see this as a pivotal moment when we look back on the advancements in AI many years from now. While this can be unnerving given the enormity of these impacts, it truly represents an extraordinary opportunity to make the world a safer and better place. And that is incredibly exciting!