The Real World Potential and Limitations of Artificial Intelligence

No longer does artificial intelligence only exist in sci-fi movies and books about dystopian futures. It’s in the here and now, continuously transforming the way in which we live and work. 

Many of us interact with AI on a daily basis - we call on Siri to give us directions to nearby coffee shops or ask Alexa to order us goods on Amazon. AI is also seamlessly supplementing and enhancing operations across a variety of industries and increasingly disrupting internal company functions. 

However, at the same time, it’s also becoming more and more apparent where AI still has limitations that prevent it from fully replicating human behavior. 

This article discusses both the real-world potential and current shortcomings of artificial intelligence, delineating between what is true today and what is still a myth.

Let’s dig in.

Advances in AI impact industries in different ways depending on the nature of the underlying processes and activities. Those that rely heavily on repetitive tasks and data analysis are ripe for disruption as modern AI can learn to recognize patterns and make sound judgments within predictable environments.

From inventory management to sales, retail companies are using AI today to support both online and brick-and-mortar operations. For example, IBM’s Watson equips online retailers with AI-facilitated order management and customer engagement capabilities. In Japan, SoftBank opened a retail location in 2016 that was staffed primarily by humanoid robots that can listen and respond to human speech. 

In the banking world, AI is augmenting both front and back office procedures. From underwriting and collection to cybersecurity and authentication, artificial intelligence is already used in many capacities and is expected to continually overtake functionality in the space. 

McKinsey & Co. recently released an analysis (see chart above) in which they calculated how much value AI could potentially create across different sectors. As you can see, the firm estimates value creation to the tune of hundreds of billions of dollars for many industries.

At the top, already discussed in this article, is the Retail space which is seeing wide scale, transformation at the hands of AI. Not only will future AI automate physical tasks, such as stocking shelves and managing checkout lanes, but it will also optimize analytics, marketing, and sales, especially with respect to eCommerce.

Second in terms of potential sheer value creation is the Transport and Logistics sector where so much time is spent processing invoices, consolidating data, and scheduling across a variety of stakeholders. Through machine learning and natural language processing, AI will be able to automate all of these activities more accurately than human personnel in less time.

Similarly, the Travel sector, which McKinsey predicts could see close to $400B in value creation, also depends heavily on logistical coordination and data analyses. And with chatbot technology improving, it won’t be long before customer service in this space is primarily performed by digital AI and humanoid robots.  

Also near the top is the automotive industry which will change significantly with AI-powered autonomous vehicles. Companies like Google and Uber are pouring money into self-driving car technology that will be able to assess driving conditions in real-time and make consistently safe decisions. With human drivers out of the equation, roads will be safer, traffic will disappear, and commuting time will be much more productive.

Overall, there is no shortage of use cases describing how AI is transforming various sectors and industries. Now, we dive into how AI is impacting internal company functions.

From talent acquisition to finance and accounting, many core processes within the average corporation will also see major change at the hands of artificial intelligence.

Many people don’t realize that most large companies today use Applicant Tracking Systems to manage job postings, schedule interviews, and screen resumes. These systems are able to sift through thousands of CVs per day and filter out unqualified candidates based on pre-programmed criteria. 

Companies are also replacing human customer service reps with chatbots that can respond appropriately to questions and address concerns. Additionally, marketing automation has exploded recently with AI leading the charge on where and when to distribute online ads based on customer behaviors on the internet.

Finance and accounting departments all over the country are also being augmented by AI that can digest massive datasets in a fraction of the time it takes human workers. Some auditing firms are even using AI to assess contracts and perform risk assessments. 

With all of the ways AI is currently changing our world, it’s easy to forget that there are still limitations with modern AI that we have yet to overcome. Below, we discuss a few of the bigger challenges facing artificial intelligence developers.

Firstly, AI requires tremendous amounts of data in order to be adequately trained to perform according to its design. In many cases, existing datasets aren’t large enough and don’t contain enough information for AI to learn how to function correctly. When this happens, humans have to spend thousands of hours labeling objects that are then fed to AI so that it can begin to build a knowledge base.

Another limitation is that artificial intelligence reflects the biases of its programmers and any bias embedded within datasets. Because AI functionality is so dependent on human intervention, it is very difficult to completely separate the two and ensure that AI isn’t created with core biases.

There is also the “black box” challenge which refers to our limited ability to decipher and understand how AI arrives at decisions and judgments. As models and algorithms grow more complex, it becomes harder to pinpoint what may have caused a specific action. As a result, it is difficult to assign accountability in certain situations.

AI also has trouble transferring learning from one experience to another, something humans are quite adept at doing. Because AI today relies heavily on predictable circumstances and recognizable patterns, it can only really function well in one type of capacity unless it is re-trained which is, again, resource intensive.

As a result of the limitations discussed above, we have witnessed a number of ways that artificial intelligence has failed to perform, some humorous and others more serious. 

Last year, Facebook shut down two chatbots, Bob and Alice, who developed their own incomprehensible language amidst a negotiation involving hats, books, and balls. The hope for the bots was that they would eventually be able to converse with humans.

In 2016, Microsoft’s Tay Twitter bot was decommissioned 16 hours after its launch as it began posting offensive content similar to what it was receiving from trolls in the Twittersphere.

Since Alexa’s launch in 2014, there have been a number of examples of the device incorrectly interpreting commands (or non-commands) from both adults and children. In January 2017, a news anchor on TV said “I love the little girl saying, ‘Alexa ordered me a dollhouse’,” which then triggered devices within earshot of the TV to also place dollhouse orders!

In Arizona, one of the more public and tragic examples of AI failure came when a self-driving Uber vehicle struck and killed a pedestrian crossing a street with her bike. After months of exciting press about how autonomous cars would change the world, the event served as a sober reminder that artificial intelligence still has a long way to go.

What Do You Think?

For most, advances in AI are incredibly exciting as the technology has the potential to truly impact our world for the better. For others, it’s hard to trust a non-human being that is designed to live and think like we do.