IBM has been banging the drum for years about the role artificial intelligence can play to support everything from cancer treatment to retail personalization.
More recently, though, IBM has started to prioritize practical advertising applications of its cognitive computing system, Watson.
Last month, IBM brought its AI to ad tech through partnerships with Xandr, Magnite, Nielsen, MediaMath, LiveRamp and Beeswax. The move followed a steady stream of Watson Advertising announcements involving AI in advertising.
In January, IBM built Advertising Accelerator, a tool that helps predict the best ads to run and tests creative versions in real time during a campaign. Over the summer came a tool that uses natural language processing to analyze social posts to match brands with relevant influencers. In September, IBM released AI-enabled weather targeting in partnership with Nielsen.
So, where does Watson Advertising go from here?
AdExchanger caught up with Bob Lord, IBM’s SVP of cognitive applications and blockchain, and the guy in charge of infusing AI into IBM’s products and services.
AdExchanger: How does IBM define AI?
BOB LORD: When we talk about AI, we mean augmenting human intelligence. We don’t intend to replace decision-making. We’re allowing marketers or supply chain managers to make decisions faster. How we do that is by crunching a bunch of data and presenting it in a way that can be used for a variety of things, from anomaly detection to deep customer analytics to powering a virtual assistant.
We also believe that AI requires transparency around the models so that people know how the decisions are being made and know the specific data that the models are learning from. Explainability – that is one of the most important parts.
Why is explainability so important?
If a business doesn’t understand how a machine makes decisions, then you’re not going to see wide adoption. Think of an insurance company that’s making underwriting decisions with a self-optimizing machine. If the business owner doesn’t know how it’s doing that, then they’re exposed. Explainability increases adoption.
And it’s also extremely important as we work with partners like Magnite, Xandr, Beeswax and others. If we allow our partners to look into our model, they in turn can go back to their other partners and say, “This is how I made this decision.” These things can’t just exist in a black box.
How exactly is IBM working with ad tech companies?
The AI technology in Watson Advertising is already being used at scale to help large enterprise brands, like HSBC, Lufthansa, Paypal, UPS and Dillard’s – so why are all these ad tech companies still trying to build this stuff themselves? It’s already been discovered. The technology is there, it’s just not been applied at scale.
We can already do real-time optimization. We’re not telling them to replace anything that they do, but we’re saying that we can add to the machinery that they have today. By inserting our Watson engine or our AI suite, they can do more with their targeting and real-time decision-making – and they have the ability to explain it.
Think of it as a layer on top of what they’ve already got that can help ensure bias mitigation for how decisions are being made – which is clearly not something that’s happening in the ad industry today.