Adapting to the Power of AI
Adapting to the Power of AI
Disruption will characterize the next stages of AI adoption.
The fact that small businesses are intentionally and aggressively adopting AI holds the first hint that the technology has enormous productivity implications. There are a number of concrete steps that can be taken in almost any business to prepare for the influx of AI. As a technology that thrives on data, collecting and storing as much information as possible in every conceivable aspect of your business can only help the eventual development and training of algorithms. Industries that have a great deal of historical data with which to train AI, such as finance, have come out ahead of those that have not systematically stored information, like medicine.
Regulation and policy prescriptions will determine growth patterns.
AI has been both enabled and constrained by coming up in a heavily regulated industry. Enabled because AI algorithms thrive on concrete rules rather than abstractions; constrained because of the inevitable concern that comes with taking human hands off the wheel. Other industries are likely to experience a similar, and unpredictable, degree of constraint as AI-driven processes are introduced. This is a key point for business leaders to absorb because the likely implications of new policies can be predicted and influenced. Liability concerns will be another factor. When AI makes a decision on the basis of its own learning, who assumes responsibility? In today’s legal environment, it will be the business that is operating the AI.
WHAT THE FUTURE HOLDS FOR AI ADOPTION. It’s an automation force multiplier that PricewaterhouseCoopers estimates will boost global GDP by as much as 14 percent by 2030
As industry adapts, AI applications will explode. Financial applications play toward the strength of algorithms in number processing, but the only thing that is slowing adoption in other industries is in finding creative ways to turn other business problems into numbers. But this can happen even in apparently unlikely or challenging situations, such as automated image recognition.
And when those problems are solved in one business silo, they can quickly leap to others. Google’s efforts at creating image-recognizing AI to build automated photo searches on the internet quickly found a use in medical imaging as a way to detect cancerous lesions or growths at far lower cost and with a much higher degree of confidence than human radiologists.
That reality neatly encapsulates one of the major challenges of AI as well as its power. When entire industries, such as medical imaging, have been built around the power of human thought and perception, there is considerable entrenched resistance to replacing it with machine algorithms.
But if efficiency will ultimately win out in those disputes, the unpredictable fragility of AI might represent new opportunities for business.
With Disruption Comes Opportunity. All AI today is considered applied AI, or “weak” AI. That’s because modern AI is built and trained specifically to conduct a single or related group of tasks. It’s considered weak by comparison to a general artificial intelligence, one that could theoretically take on any problem solvable by intelligent thought, but that also makes it weak in comparison to general -- natural intelligence -- the kind that humans use.
The human ability to pattern match and apply experiential learning is not limited to specific situations in the same way as a weak AI. When problems appear outside of the context in which it was trained, a weak AI will need to rely on human recognition and assistance in resolving those problems. That will represent a new business opportunity.
It’s more difficult to see exactly where these opportunities will emerge, but some hints can be found in current AI implementations.
One of these revolves around the push toward implementing self-driving cars. A classic application of weak AI that is being pursued by companies from Uber to Google, algorithms can process sensor inputs and use rules and experiential training to react faster and more safely to road hazards than any human driver. AI won’t get distracted by a cell phone or forget the speed limit.
But it is learning only under the slow and careful tutelage of old-fashioned, slow-witted human beings. While self-driving cars report an impressive safety record for miles driven, all of them have done so with human supervision behind the wheel. And interventions are common where algorithms misinterpret a reading or make the wrong logical assumption about road conditions.
This combination of human and AI is what has earned that exemplary safety record, not one or the other. Companies that find ways to combine the strengths of AI algorithms with the strengths of human creativity and insight are likely to reap the greatest benefits of AI.