"Artificial intelligence (AI) could spell the end of the human race,” said Stephen Hawking when the BBC asked him what he thought of AI. Unsurprisingly, sentiments like these from thought leaders — along with countless movies in which evil robots take over the world — has cast a long shadow over the understanding of AI.
So much so, in fact, that at the Mindshare Huddle on 15 November, much of our discussion focused on the public perception of AI. Now, I’m not going to argue with Stephen Hawking’s ideas on what AI might become. But the reality of AI today, and for a long time to come, is much less dramatic and more benign.
AI Already Walks among Us
In fact, a Deepmind AI simulation even taught itself to walk. AI is here and you’re probably even using it while you read this blog post. Apple’s Siri speech recognition, Netflix’s recommendation engine, Google’s search algorithms — all AI we use today and take for granted. And very soon, they’ll be joined by AI’s that drive cars, revolutionise medical science and help us in lots of other ways.
These are examples of a specific type of AI, called machine learning. Machine learning uses a computer’s ability to process large amounts of data, discover patterns in that data and then turn it into an output that serves as an instruction or command to machines. The technique has been around since 1959, but today’s powerful computing processors and ever-increasing data storage capacity allow it to be used far more widely than ever before.
How Does Machine Learning Work?
Machine learning calculates the probability of a future event based on observations from past data. Let’s say I wanted to predict whether I am going to take my two kids to a playground today.
I’d feed my machine-learning algorithm with lots of data – the weather, the day of the week, the school holiday calendar and so on — about all the days in the past when I had and hadn’t taken my kids to the park. AI would then sort this data, to spot the variables common to the days on which I took my daughters to play on the swings.
I’m fairly certain that it would assign the highest probability scores against weekend-days with sunny weather and the lowest probability scores against days that were both work-days and school-days on which it was raining.
But who knows? One of the great things about AI is that it can and often does spot patterns that we can’t, patterns that can help us improve our businesses and our lives. And the more data points and data volumes we give to these systems, the better their predictions get.
So How Does All This Apply to Advertising?
Serving digital ads produces vast amounts of data points and machine learning is uniquely suitable to learn from this data and place a specific value on a programmatic ad impression the DSP offers based on observations from past data. There’s also far more inventory now than you could possibly access by place using manual insertion orders.
A human campaign manager can look at a relatively small set of variables: the time of day, publisher and so on which have performed the best. AIs such as Xaxis’ Copilot can look at thousands of variables, in near real time, to find exactly the combination of targeting variables for maximum performance.
But that doesn’t mean that the role of campaign manager is obsolete. In fact, it’s more important than ever. For the AI to deliver the right output, it needs the correct input. That requires knowledge of the client, the brief, the industry and the know-how on how to integrate client data sets to ensure that Optimisation is as closely aligned as possible to the Outcomes the advertiser aims to achieve.
Success also depends upon campaign managers who know the AI well enough to work with engineering teams to develop new data-led strategies. At Xaxis, our campaign managers even work with data scientists and engineers to develop new, innovative campaign-specific algorithms for Copilot.
AI-powered systems in advertising are already invaluable tools to make media investment go further and ad campaigns perform better. In the future, they’ll be even cleverer, more adaptive and higher performing. And that’s great news for advertisers but also for those of us working in the ad industry.
