The following is a guest contributed post by Blis President Harry Dewhirst.
Current AI systems use training algorithms and brute-force computational strength, as opposed to real intelligence. If an alteration is made to a situation involving AI, it will currently not adapt as it has no ‘understanding’ and therefore cannot recognize new strategic patterns. There’s no doubt that future technologies such as AI and the Internet of Things (IoT) will be put to work properly by marketers, but the questions remain as to when and what it will take.
The limitations of algorithms
Algorithms are limited in their learning to a set of defined tasks, so if a self-driving car approaches a zebra crossing with some pedestrians standing nearby, it will stop and wait for them to cross, even if they have no intention of crossing, and won’t respond to their uniquely human gestures.
The limit to machine learning is data; with new streams constantly feeding in from all directions, there is an opportunity to learn and discover, but also a major technical challenge. As the volume of new devices like hearables and wearables feeding into the IoT rapidly expands, they provide a treasure trove of data which needs to be speedily and accurately analysed to allow the IoT to live up to its potential. However, the manipulation of Microsoft’s Tay chatbot, which learned to be racist and rude in less than a day, reflecting the information it was fed and the behavior that it mirrored, shows us the importance of being able to logically sift good and bad data inputs.
Extracting meaningful insights
Machine learning can help companies take the billions of data points they have from IoT devices and boil them down to what’s really meaningful, so that better decisions can be made. This is essential to find patterns, correlations and anomalies, since current human approaches of data analytics don’t scale to IoT volumes.
Although location data from connected TVs (CTVs) and other IoT devices that rarely move currently has limited value, smaller mobile devices provide one of the largest location data sets as they are incredibly dynamic by nature, with mobile apps in particular proving very useful input data tools, feeding CTV advertising optimization.
The importance of context
As advertisers build in-depth profiles of users based on data from IoT devices, the importance of context to advertising is increasing. There needs to be a good reason to reach a consumer in their home and, while artistry and storylines are important in engaging audiences, the context in which ads are delivered is today equally as important to engagement as the uniqueness of the ad storyline and copy itself. Based on data insights, personalized ads can now be delivered to the right person, at the right place and time to ensure the message engages and resonates, creating a return on investment. Intelligent analysis of location data is critical because the “where” can significantly influence the choice of campaign served to a particular user at a given time.
Although the volume of data and the challenges and insights presented by our increasingly digital, mobile-centric world have made the job of advertising to consumers more difficult and multi-layered in some ways, if the correct infrastructure is in place, it is extremely helpful. Depending on how these advances are implemented, ads will either cause concern as they become invasive and affect overall privacy or, if they work smoothly, consumers will come to expect them more and more as they seamlessly fit in their lives.
While machine learning is enhancing the capabilities of the industry, real AI is expected to bring further advances; precisely how it will achieve this is hard to predict, but it is likely to generate surprising insights into customers’ requirements.
