It's Time to Rethink "Artificial Intelligence"

What is 'intelligence' and can it be 'artificial'? From Merriam-Webster:

Intelligence(n): 1. the ability to learn or understand or to deal with new or trying situations

2. the ability to apply knowledge to manipulate one's environment or to think abstractly as measured by objective criteria (such as tests)

Granted, subsequent M-W definitions for 'intelligence' do reference computer algorithms; however, I believe that 'intelligence' should be considered, at least in part, as a measure of adaptability and creativity, not just the ability to perform certain functions well. This is a distinction that is lost in the current dialog regarding Artificial Intelligence, I think largely to keep the conversation simple -- though the devil is, as they say, in the details.

Surfing the Hype Cycle

I would argue that, in it's current instantiation, 'Artificial Intelligence', while certainly artificial, is not 'intelligent' in any meaningful way. Rather, the universe of technologies and capabilities that fall under this umbrella is composed wholly of highly specialized routines, each performing one specific function really well. Examples -- Google's auto-drive AI is exceptionally good at navigating streets, but would be horrible at piloting a boat or an airplane; IBM's Watson is perhaps more flexible in concept, but still is composed of individual, specialized functions that separately perform activities like natural language processing, feature detection in images, crushing humans at Jeopardy, etc.

We are clearly riding the top of the wave of excitement (hysteria?) in terms of what AI can do and should do. This is great from the perspective of building awareness and enthusiasm for new, innovative capabilities, but harmful in terms of the psychological impact these pumped-up perspectives on AI capabilities has on people as individuals. Pushing AI as the 'obvious solution' for practically any problem will lead (and has led) to wild accusations and headlines such as 'In Two Years, AI Will Take Over Every Job in the X Industry' or 'AI is Taking Over the World to Rule Humankind as Our Machine Overlord.' These exaggerations sell ads for sure, but also lead to irrational concerns and fears.

A Call for Reason

My intention is not to diminish or devalue the new capabilities that have been enabled through AI. On the contrary, solutions based on current AI concepts and techniques can perform amazing and important functions -- from accelerating the discovery of new medical diagnostics and treatments to providing better security solutions through pattern analysis and fuzzy threat detection. Proper application of AI and related machine learning techniques has led to significant strides in many fields and is leading us toward huge improvements in economic conditions and overall quality of life.

It is extremely important to note, though, that advances in AI so far have been fairly modest; only in the last 7-8 years has computing and storage capacity been sufficient to enable real-world application of rule-based and neural network methods to solve interesting problems in real-time. While strides in some areas have been rapid and significant, progress on other fronts remains slow. In truth, our depth of understanding of AI as a concept particularly neural networks and deep learning, is limited. Although the field of neural networks was launched in the 1950's, much of what we know and apply today is still discovered through trial-and-error (the choice of a specific neural network architecture to address a particular problem is still mostly an educated guess). Universities, companies, and governments are investing significant time, effort and money to improve our understanding and ability to construct better AI, but moving this capability from the category of 'art' to 'science' will consume data scientists and data science resources for quite some time to come.

Again, this fact does not detract from current progress or intentions, it merely demands attention to ensure expectations on the whole are reasonable. Expecting AI suddenly replace large portions of legacy technology, or displacing significant numbers of people for that matter, is unrealistic and dangerous.

The Future is... Soon

All this being stated, recent developments in a number of heretofore unrelated technology fields is showing significant promise in elevating AI beyond its current form. In particular, combining concepts of AI with Blockchain is a field of research and product development worth watching. The goal in bringing the two technologies together is to enable truly automated and fault-tolerant processes that can be applied broadly to improve productivity across many different industries and functions. This convergence has the potential for changing business and our society at the local and global levels.

One concept that is receiving much attention now, referred to as 'Radical Decentralization', is aimed at improving how businesses and communities function through tighter integration of services and fault-tolerance through shared infrastructure -- leveraging Blockchain to provide security and custody of data, the Internet of Things to seamlessly track and monitor resources, and AI to coordinate needs dynamically will converge to support local community resilience while tightly integrating with the global economy for stability.

And, Now... So What?

At the end of the day and even in its current form, AI is pretty awesome; it has made and will continue to make our lives better in many ways. It is just so important, given the current hype surrounding AI, that we facilitate a more balanced and realistic view of the current landscape of capabilities, as well as foster more modest expectations for what is possible and reasonable today -- especially in the public domain. A rational perspective on the true impact of AI on society will greatly reduce concerns of personal impacts and minimize economic and social fears.