Most businesses run into the same problem when they begin implementing AI:
Where do you get all the data?
The answer depends entirely on the business and its available resources. It's why we do tailored consulting with brands on the best AI approach for them.
And it's why we found this story about how Facebook gets data to train its own AI particularly insightful.
(Stick with us: Even if your company doesn’t have the size and clout of Facebook, this has lessons for you.)
Turns out, Facebook used Instagram photos to train AI to recognize objects. Facebook analyzed 3.5 billion Instagram photos and their hashtags. The hashtags taught AI algorithms how to categorize photos. After enough learning, the AI could then categorize photos itself.
One big benefit: Instagram users had labeled the data for free, thanks to the hashtags they used on each post. Says WIRED, “It provided a way to sidestep having to pay humans to label photos for such projects."
This whole scenario provides lessons for brands of any size that want to use AI.
Lesson 1: Make sure you have data.
AI tools usually need large amounts of data to “learn.” And that data has to come from somewhere.
Too many brands we speak to think AI implementation is just a matter of picking a tool and flipping a switch. That's understandable: AI gets talked about a lot like it’s some type of magic wand you wave and it “works.” The reality is usually messier.
As Facebook's story shows, you need a lot of data to train AI systems. Facebook was able to use a huge publicly available dataset to train its AI. Your brand probably won't be so lucky.
The first question to ask is:
What data do we have available?
This is why guidance during this process is critical. Deep vendor research, tool recommendations, and an AI roadmap are must-haves. Only with them can you determine exactly how much data you need and what type of data you need. And, if you don't have the right data, you need a plan on how to get it.
Lesson 2: Make sure it’s good data.
Even if you have data available, it needs to be good data. Data must be cleaned, formatted, and labeled properly. Make no mistake, this is not an easy housekeeping item, it's an in-depth technical process all its own.
Facebook found an elegant way to label its data. Social media users did it for them. Don't expect that to be the case with your data. It will likely need to be formatted and labeled by humans. That takes time and money.
Train yourself out of magical thinking when it comes to AI:
The technology is sophisticated, but it's not sorcery. AI is like any other machine that requires a specific type of fuel. You have to put in the right stuff in the right amount for it to work.
You can bet Facebook spent a lot to make sure the Instagram data the algorithms used was using good data.
Lesson 3: Make sure you have enough resources for AI.
Getting the right data and prepping it require resources: money, time, and the right talent. Too many brands don't truly understand the resources required for AI.
“Facebook’s project also illustrates how companies need to spend heavily on computers and power bills to compete in AI. Computer-vision systems trained from Instagram data could tag images in seconds, says Paluri. But training algorithms on the full 3.5 billion Instagram photos occupied 336 high-powered graphics processors, spread across 42 servers, for more than three weeks solid.”
Now, these are the resources you need if you're Facebook and you're building your own AI. Your brand will likely start at a much smaller scale.
But it's useful to start thinking about the resources required to build your own AI. It's short-sighted to plan an AI strategy with the expectation you'll build it all in-house. In fact, many brands will get much further, much faster with out-of-the-box solutions. They might even do well to see how their existing tools use AI—and start there.
In any case, we can't emphasize enough: You need a clear, sensible AI strategy before you start adding AI to your tech stack.
We offer guidance on how to build your strategy with our subscriber list, so sign up here.
Also, if you have formal AI consultation needs, we can help if you get in touch here.
As Chief Content Officer, Mike Kaput uses content marketing, marketing strategy, and marketing technology to grow and scale traffic, leads, and revenue for Marketing AI Institute. An avid writer, Mike has published hundreds of articles on how to use AI in marketing to increase revenue and reduce costs. Mike is the co-author of Marketing Artificial Intelligence: AI, Marketing and the Future of Business (Matt Holt Books, 2022). He is also the author of Bitcoin in Plain English, a beginner’s guide to the world’s most popular cryptocurrency.