The Skeptic's Guide To Assessing Artificial Intelligence

Post written by

Jeffrey Ton

SVP of Product Development and Strategic Alliances at InterVision, and author of "Amplify Your Value: Leading IT with Strategic Vision."

Getty
Humans have a tendency to race toward the next big breakthrough. In the 1960s, we were part of a race to the moon. The early 2000s was a race for the best consumer electronics. Now we are on a race towards true artificial intelligence (AI).

However, the desire to be first and offer the most advanced technology has led to the unfortunate trend of AI washing. AI washing is when vendors label their technology as “artificial intelligence” when it is not true AI, leading many businesses to quickly become skeptics of any AI technologies. You see, all AI is machine learning, but not all machine learning is AI. To help mitigate this skepticism, I have outlined how businesses can distinguish between simple machine learning and actual AI capabilities, as well as how to vet providers for true AI. 

The Key Components Of True AI  

One of the primary reasons that many business leaders are skeptics of artificial intelligence is that they do not fully understand how to differentiate between machine learning and AI. Machine learning is pretty simple. This type of technology consists of a series of algorithms that use an input dataset along with known outputs to learn underlying patterns. This allows them to make predictions on new input data.

On the flip side, AI is made up of machine learning, but the outcomes are no longer defined and the possibilities are unlimited. For example, machine learning algorithms are typically used as a training set of data to teach a system initial commands, but once an environment starts changing, AI will tap into capabilities similar to human intelligence and make decisions on its own. When AI is in play, a much more complex system is in place that likely produces more value for an organization.

Today, AI implementations are defined as narrow AI, meaning they can perform a specific task very well -- in most cases, more accurately and more efficiently than a human. 

How To Test A Provider For AI 

Once a general understanding of true AI is established, the only skepticism remaining typically stems from the fact that businesses looking to implement some type of AI may not trust providers to be honest about their products' capabilities. It’s easy to say that a product has AI capabilities, but it's much harder to put true AI into practice. Since true AI will learn and become smarter over time, it’s important to ask providers leading questions to determine if their technology has this capability. A good place to start is by focusing on these six key questions: 

1. How does your product improve over time?

2. What decisions can technology make and adapt to over time?

3. What’s the feedback loop for the AI engine to learn?

4. Does it need human feedback? 

5. What will my company be able to do with this AI engine?

6. How is this product going to help my human workforce make better, more informed decisions?

If the provider cannot effectively answer these questions, the system is most likely utilizing machine learning algorithms and not true AI. 

It’s also important to pay attention to the type of staff the provider employs and who would be handling any technology implementation or monitoring. A company that is implementing true AI will have data scientists on staff or will be outsourcing to an organization with data scientists, not just hiring programmers who can build algorithms. This is because building an AI model is incredibly complex and a data programmer will not have the right skills to build a program that is both adaptable and able to gain knowledge over time for informed decisions. 

Taking a pragmatic approach to AI will not only serve in selecting a true AI solution, but it will also help to enable (and challenge) your company’s long-term strategy. AI as an innovation may still be in its infancy, but it won’t be going away. Like cloud technology, it will eventually consume more and more of the market and therefore continue to transform customer expectations. Knowing this market direction and getting buy-in from company leadership for how to shift in this direction is a valuable task to begin now.

Forbes Technology Council is an invitation-only community for world-class CIOs, CTOs and technology executives. Do I qualify?
">

Getty
Humans have a tendency to race toward the next big breakthrough. In the 1960s, we were part of a race to the moon. The early 2000s was a race for the best consumer electronics. Now we are on a race towards true artificial intelligence (AI).

However, the desire to be first and offer the most advanced technology has led to the unfortunate trend of AI washing. AI washing is when vendors label their technology as “artificial intelligence” when it is not true AI, leading many businesses to quickly become skeptics of any AI technologies. You see, all AI is machine learning, but not all machine learning is AI. To help mitigate this skepticism, I have outlined how businesses can distinguish between simple machine learning and actual AI capabilities, as well as how to vet providers for true AI. 

The Key Components Of True AI  

One of the primary reasons that many business leaders are skeptics of artificial intelligence is that they do not fully understand how to differentiate between machine learning and AI. Machine learning is pretty simple. This type of technology consists of a series of algorithms that use an input dataset along with known outputs to learn underlying patterns. This allows them to make predictions on new input data.

On the flip side, AI is made up of machine learning, but the outcomes are no longer defined and the possibilities are unlimited. For example, machine learning algorithms are typically used as a training set of data to teach a system initial commands, but once an environment starts changing, AI will tap into capabilities similar to human intelligence and make decisions on its own. When AI is in play, a much more complex system is in place that likely produces more value for an organization.

Today, AI implementations are defined as narrow AI, meaning they can perform a specific task very well -- in most cases, more accurately and more efficiently than a human. 

How To Test A Provider For AI 

Once a general understanding of true AI is established, the only skepticism remaining typically stems from the fact that businesses looking to implement some type of AI may not trust providers to be honest about their products' capabilities. It’s easy to say that a product has AI capabilities, but it's much harder to put true AI into practice. Since true AI will learn and become smarter over time, it’s important to ask providers leading questions to determine if their technology has this capability. A good place to start is by focusing on these six key questions: 

1. How does your product improve over time?

2. What decisions can technology make and adapt to over time?

3. What’s the feedback loop for the AI engine to learn?

4. Does it need human feedback? 

5. What will my company be able to do with this AI engine?

6. How is this product going to help my human workforce make better, more informed decisions?

If the provider cannot effectively answer these questions, the system is most likely utilizing machine learning algorithms and not true AI. 

It’s also important to pay attention to the type of staff the provider employs and who would be handling any technology implementation or monitoring. A company that is implementing true AI will have data scientists on staff or will be outsourcing to an organization with data scientists, not just hiring programmers who can build algorithms. This is because building an AI model is incredibly complex and a data programmer will not have the right skills to build a program that is both adaptable and able to gain knowledge over time for informed decisions. 

Taking a pragmatic approach to AI will not only serve in selecting a true AI solution, but it will also help to enable (and challenge) your company’s long-term strategy. AI as an innovation may still be in its infancy, but it won’t be going away. Like cloud technology, it will eventually consume more and more of the market and therefore continue to transform customer expectations. Knowing this market direction and getting buy-in from company leadership for how to shift in this direction is a valuable task to begin now.

Forbes Technology Council is an invitation-only community for world-class CIOs, CTOs and technology executives. Do I qualify?

Jeffrey Ton is SVP of Product Development and Strategic Alliances for InterVision, an author and a speaker. 

Forbes Technology Council is an invitation-only, fee-based organization comprised of leading CIOs, CTOs and technology executives. Find out if you qualify at forbestech...