A few days ago, the internet nearly broke itself with the circulation of this story (or some version of it):
Facebook AI Research team shuts down the program after AI Robots secretly invent their own language
Before jumping to “The Terminator” and “Westworld,” remember that we are still a ways from “Skynet” and the media loves to sensationalize – even though the real story is every bit as intriguing for marketing leaders hoping to implement AI in their 2017 marketing plans.
So, what really happened?
A couple of months ago, a blog post by FAIR (Facebook Artificial Intelligence Research) described a dialog between two AI agents at Facebook as being peculiar and interesting. Essentially, the two agents began to diverge from plain old English into “seemingly nonsensical sentences” because of a programming error. All of this seemed straightforward, garbage in garbage out. However, the oddity was when the two agents, speaking in what seems like gibberish to us, actually understood each other. Since the purpose of the work done by FAIR was to improve human-AI communication, the error was fixed and the team moved on.
Lucky for us, this myth was easily debunked but unfortunately not all AI myths are as easily dismantled. Myths are consistently holding companies back from fully taking advantage of the advancements in artificial intelligence. Here are a few:
Myth 1: A team of data scientists is required to deploy AI.
Too much hay has been made around AI which sometimes translates to a lot of misinformation. Not unlike our Facebook story, people look at AI as a self-fulfilling prophesy from watching too many science fiction movies. We immediately assume that for someone to play in the AI sandbox, they must carry a PhD in AI or HCI.
Don’t be afraid to get involved in AI activities without a background and few degrees in artificial intelligence. Deploying successful artificial intelligence automation and capabilities requires every team in both marketing and IT, and especially the marketing leaders to have a vision of what problem AI will solve. This includes the data layer analysts that must ensure data integrity as well as business analysts and strategists to formulate a vision on how to utilize AI. There are also the functional resources who are hands-on with the tools and not forgetting the marketers that interpret the data and produce insights. There are operations resources that maintain the infrastructure where AI can function and development teams that integrate these tools together. And on and on. The AI sandbox is open to the public without invitation.
Myth 2: AI taking our marketing jobs.
Just as experienced with the industrial revolution hundreds of years ago, the AI revolution will not cost us jobs but rather evolve those jobs. The jobs that will eventually wither away are those that require manual effort, like data creation, curation and analysis. However, new jobs will emerge that require more critical thinking. While AI can provide analysis and insights into a set of data, it takes a human being to absorb those insights and make recommendations because we have context.
For example, if an AI agent analyzes the searches done by users on a website, it can do that and even provide insights based on that analysis. An insight in this instance could be, “Users search for product names more often than product numbers, brands or categories.” Such an insight is critical because it highlights user tendencies when they interact with your website. However, what to do with this insight is an even more critical task that belongs to humans. A human should weigh the options and make a recommendation. Options include:
1) Don’t include product numbers, brands or categories because of performance, especially since users don’t often include them in their search queries.
2) Market brands, categories and numbers better so users are aware of them and utilize them more in search queries.
3) Recreate your product taxonomy to follow names over numbers, brands and categories.
#3 seems pretty far-fetched, #2 seems reasonable for brands but not categories or numbers, and #1 seems exclusionary and could backfire for those users with only a product number. A human will most likely find a compromise and recommend a combination of these three and others that I didn’t think of. But at the end of the day, only a human can do that.
Myth 3: AI will need a few years to be ready for the enterprise.
A very common myth around AI is that it is too costly, too time consuming and is designed for big problems. This cannot be further from the truth.
AI can be used for all sizes of problems, specifically ones that can benefit from automation and iteration. Back to the search problem from Myth #2, One can create an AI algorithm that utilizes machine learning and improves its search parameter based “context”. So over time, the AI can learn which page do users usually use the product number to search from versus the product name or category to eventually improve both search results relevancy and performance because of ongoing evolution.
Coveo, for example, is an enterprise-ready solution that utilizes machine learning as part of its product core. Don’t be afraid of how artificial intelligence can make a difference for your marketing goals – just make sure that your team is laser focused on your goals and selecting the solution with AI that can meet those. Find out more about Coveo on their site to see machine learning in action.
Over the past 17 years, we have completed thousands of digital projects globally. We have one of the largest and deepest multi-solutions digital consulting teams in the world. Our proprietary processes and years of Digital Experience expertise have earned us a 97% customer satisfaction rating with our clients ranging from Global Fortune 1000 to Mid-Market Enterprises, leading educational institutions, and Non-Profits.
Contact us today if your organization needs assistance with your Enterprise Search as well as content, commerce and digital marketing.