BETA
This is a BETA experience. You may opt-out by clicking here

More From Forbes

Edit Story

The Next Stage In The Algorithm Revolution: Creating Artificial Intelligence With A Human Touch

Forbes Technology Council
POST WRITTEN BY
Raj De Datta

For brands seeking to capitalize on the power of artificial intelligence (AI), it’s the best of times to use algorithms. It’s also the worst of times. 

The best: the most effective and popular AI algorithms -- such as those used in basic search, computer vision and natural language – are now available in the cloud from companies like Amazon, Google and Microsoft. This has transformed an exceptionally valuable but often limited innovation into an accessible and democratized technology: AI algorithms “as a service.” The worst: this “commoditization” of AI can greatly diminish the likelihood of creating unique solutions, customizing to specific use cases -- and offering a competitive advantage.

The massive machine learning (ML) infrastructure needed to develop these algorithms is readily available in the open-source world or via APIs in the cloud. This means if you’re a brand looking to enhance the customer experience through an AI algorithm designed for personalization, optimization, search or recommendations, you can find them online -- but so can your competitors. 

In other words, how do you differentiate the way you target customers to sell your brand?

Customizing The Commoditized Algorithms

The reality is that commoditized, readily available algorithms don’t generally provide a competitive advantage.

As the CEO and co-founder of a company that offers AI search and merchandising solutions, I believe that’s where customization comes in and brings us to the great irony of artificial intelligence today: Without human intelligence to customize algorithms, AI is just another commoditized tool. It's an extraordinarily helpful one, to be sure, but nearly not as valuable as it could be. It’s time to join human expertise with technology to program algorithms.

The fashion industry provides a useful example. The hottest trends drive sales for many companies, so algorithms that learn from data around the past will recommend outfits -- often with great success. But they won’t always feature newer trends that don’t yet have a track record or pattern to prove their value. 

In other words: today’s algorithms are good at learning from the past -- but to propel a company forward and create truly inventive consumer experiences, I believe companies need to provide a fresh look into the future with purpose-built algorithms they create with human insights.

If you can customize on top of your algorithm to map it to a specific application -- such as recommending a side dish to order at a chain restaurant based on the diner’s history and interests, as well as their main course -- then you can enhance that consumer’s experience while promoting another sale. In other words, I believe better experiences require more than AI-powered recommendations; companies should also customize their algorithms around their use case. And in many cases, today that requires the human touch. 

The Bias Problem

Another “elephant-in-the-room” issue gaining traction is the bias inherently associated with AI in general and commoditized algorithms in particular. According to Deloitte, media stories related to AI and ethics doubled in 2018 over previous years, and approximately one-third of executives surveyed by Deloitte earlier this year said the ethical risk of AI is a top concern.

I believe it's a serious concern. Many datasets lack diversity -- and many algorithms fail to understand and parse biased data. For example, if you’re searching for an SVP of marketing on a networking website, the algorithm will likely look at you, your connections, where you live and your background. Recommended candidates may lack diversity; instead, they could look a lot like you and your existing networks. 

You can create an ideal, customized algorithm in this situation that adds diversity by aggregating and analyzing other attributes, such as gender and background, then looks at the resulting data sets to make sure there’s at least a sprinkling of diversity in those results. 

The chase for useful personalization shouldn’t sacrifice data diversity, which I believe is critical for driving superior business outcomes. To break that vicious cycle, it’s up to humans to stay involved and adjust algorithms to welcome outside influences and integrate them into our thinking.

Human Intelligence Only Strengthens AI

Amid the AI revolution, many technology companies are racing to harness AI, ML and big data and drive greater customization and personalization into individual customer experiences. They can not only account for current consumer needs, but also work proactively to anticipate the tastes, moods, desires -- and problematic issues -- of their customers. 

There are three steps you can take to make this happen in your own algorithms:

1. Understand and define the business objectives you want to achieve with your AI. For that to happen, and for IT teams to take advantage of the software, implement programmatic or visual tools to see how the AI is working in its present state.

2. The second step is to apply the highest levels of transparency to the AI. IT leaders and programmers should be able to explain what the AI is doing to customize the AI for their specific business purposes. Analytics is a vital part of the equation to understand what’s working in the algorithm and what’s not.

3. Then, you can map your business to the algorithms by programming your own intellectual property into them. Essentially, you’re tuning the AI specifically to your business needs. That’s the best way I know to maximize your competitive edge. You can look at any algorithmic output and see whether it’s aligning with your business objectives.

Ideally, sometime in the not-too-distant future, a non-technical person will be able to focus on real business outcomes in these AI algorithm tools by editing or augmenting the outputs and won't have to rely on a technical person to do it. This combination of human and machines working seamlessly together will require trust from the very top, which you can more easily achieve if your organization gathers evidence and validation that its algorithm customization is working.

Algorithms are getting smarter every day, but they can’t do it alone -- yet. As algorithms continue to evolve, look for humans to play a pivotal role sharpening their intelligence, connecting the dots and keeping us marching toward a smarter future for commerce and beyond.

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