Sanatu Kolay, SVP of engineering at marketing platform Turn, recently penned an article with the provocative headline “AI in Ad Tech Mostly Fiction”. In it, he suggests that the kinds of tech used in ad tech, even in its most sophisticated forms, are “not really AI”.
This is a tired misconception. I remember studying in the field of Artificial Intelligence at Stanford and MIT, working with or taking courses from pioneers of the field including Nils Nilsson, Marvin Minsky, Terry Winograd, and John McCarthy. The period was the 1980’s, and AI was going through what they called at the time the “AI Winter”. Nobody could get funding for companies that employed AI technology, because it was a “dirty word”. AI had been hyped in the 1960’s and 1970’s, as rapid progress was made — it was proclaimed widely that “within 10 years” we would have human-capable AI that could pass the Turing Test and possess all manner of other smart capabilities.
By 1985, that had not happened, and the world said “enough is enough”. Like the boy who cried wolf, AI researchers were shunned with scepticism at any claims they continued to make, and they retreated into their academic offices to make quiet progress.
But in the intervening 30 years, something fascinating has happened. Computers have become exponentially faster and more capable, following Moore’s Law. AI algorithms for machine learning, planning, optimisation, and reasoning have improved too. Networks of powerful computers can now work together to solve problems, taking advantage of network effects as described by Metcalfe’s Law.
However, because the world has been burned before, it was not until just a few years ago that the academics came back out of their offices to be willing to call the latest round of work “AI”.
In reality, AI has been making steady, and now rapidly increasing, progress the whole time. The fact you can “Google” anything and get relevant results counts as AI. The path planning built into your car to route you to your destination uses algorithms developed by AI researchers. Speech recognition, face recognition in Photos apps, even optimisation of how packages should be routed to you through courier services all rely on advances in AI. And the new generation of Deep Learning is only making these systems richer and better, as well as enabling new previously unheard-of capabilities like self-driving cars and intelligent personal assistants.
Ad Tech is no different. While most companies working in ad tech have concentrated on providing easy-to-use workflow for human operators to have more control over where ads are shown, a few have really concentrated for years on applying AI techniques to the problem of optimising advertising performance. The vast amount of data available about individual consumers, together with the proliferation of opportunities to show personalised ads and the shift of consumer attention into digitally addressable media have all conspired to make this possible.
While a few years ago it would have been fashionable to dismiss these advances as “not really AI” because they fall short of passing the Turing Test, to do so today is laughable. One would also have to exclude self-driving cars, speech recognition systems, and all the other advances which we enjoy that came out of the same technological heritage, and which came from researchers trained in AI, who consider themselves to be working in the field of AI.
If the standard instead is “a fully automated system that requires no human intervention”, I would agree that designing the ads to elicit an emotional response is (for now) still the domain of clever, creative humans. But the task of choosing who should see what ads, based on what kinds of people and contexts (moments) have led to good results in the past, is well within the domain of full automation today.
Instead, let’s embrace the amazing capabilities that AI-powered ad tech brings to the challenge of effective marketing, and move forward to new initiatives such as “how can AI and human operators work together to achieve even better results than either could alone?” and “what elements should I consider adding to my clever creatives to make them even more impactful on certain groups of consumers?”
By Mark Torrance, chief technology officer at Rocket Fuel
PrivSec Conferences will bring together leading speakers and experts from privacy and security to deliver compelling content via solo presentations, panel discussions, debates, roundtables and workshops.
For more information on upcoming events, visit the website.
comments powered by Disqus