Anticipating an Artificial Intelligence Arms Race

As part of my role I get to participate in several meetings regarding the ethical implications of AI. We all know that AI is a combination of statistics, math and computer science. However, now with large amounts of training data as well as the scale of cloud platforms and productivity, we are seeing more than the thawing of the historical AI winter. In many respects I'd think we are currently witnessing an explosion of AI capabilities at an increasingly exponential rate. 

 In my article "Is Artificial Intelligence Truly Intelligent" found at https://www.dhirubhai.net/pulse/artificial-intelligence-truly-intelligent-eric-charran/, I discussed the difference between weak and strong (AGI) AI which is best depicted in the graphic below.

As more organizations turn to master and democratize weak AI and integrate it into applications and digital experiences, it stands to reason that these same purveyors of AI are going to turn to invest in combining weak AI with Strong AI (an artificial general intelligence). This combination unites weak capabilities together with cognition to rival or exceed human abilities. 

 Off to the Races - How it Begins

 While many AI organizations that are conducting this investment and investigation have focused on adopting an ethical substrate or partnering to share a common one, it stands to reason that there are bad actors in the world that choose not to be constrained by this type of ethical standard.  In this scenario, we must imagine bad actors, either political, criminal or otherwise, that choose to leverage AI in situations where the major organizations that are investing in AI have decided and agreed not to. 

 One scenario is if a bad actor, as in a state or political world power, that ignores a common ethical substrate for AI chooses to create AI-based weaponry (both weak and strong AI), it stands to reason that other entities will need to develop the same or greater technology as a deterrent or counter. This brings about a scenario where an actual AI arms race may manifest. Imagine a case where AI-based weapon systems are developed and built into combat weaponry which increases the effectiveness of a military to the point where a non-AI enabled opposition force would be at a significant disadvantage. The observation quickly becomes apparent that in order to compete and maintain parity in terms of effectiveness and force that the opposition military would have to invest in their own.

 Innovation for Competing Capability eats Ethics for Breakfast

 This is significantly perilous because as we've seen the rate of innovation and the rapid democratization of these capabilities far outstrips our ability to ask "we know we can develop these capabilities, but SHOULD we?" In an arms race, the question becomes "we know our competitor or adversary is developing these capabilities, but how can we NOT develop our own?" The danger in this line of thinking is that in a rush to develop and implement AI capabilities in systems of warfare for example, the ethical considerations are secondary to matching those of an potential adversary. 

 The danger is that we are more likely develop AI capabilities that are harder to control, monitor and regulate in a competitive scenario with nation states or potential military adversaries. A race breeds the scenario where the need for innovation to match adversary capabilities outmatches ethical considerations. And I think we've all seen how that movie goes.

 Digital Geneva Convention

 This makes the need for a digital "Geneva Convention" all the more salient. A code of conduct and enforceable regulations that mirror an already established template following weapons of mass destruction may be in order. While no one wants to see hard regulations on the usage of AI to stifle ingenuity, human augmentation and productivity, we do need to make sure that humans have an agreed up on ethical north star. With the advent of any new technology or capability, the threat of misuse, misappropriation and abuse represent a new way for humans to collaborate, agree and move forward with care.

This is something that Brad Smith, President and Chief Legal Officer at Microsoft has championed for some time now. For details, watch the video and read the post at this link: https://blogs.microsoft.com/on-the-issues/2017/02/14/need-digital-geneva-convention/