In 1955, computer scientist John McCarthy coined the term artificial intelligence. Just five years before, English Mathematician Alan Turing had posed the question, “Can Machines Think?” Turing proposed a test: could a computer be built which is indistinguishable from a human? This test, often referred to as the Turing Test, has sparked the imagination of AI researchers ever since and been a key idea in the field.
In the late 1990s artificial intelligence made its mark again, when IBM’s Deep Blue beat the world chess champion Gary Kasparov. Since then, advances in computing power and data accumulation have led to a proliferation of new technologies driven by artificial intelligence. From self-driving cars to self-regulating thermostats to image recognition, artificial intelligence (AI) is proving to be one of the most transformative technologies of the century.
For the data laden marketing industry, AI has become more relevant than ever. AI offers the promise of analyzing millions of disparate data points, proving out creative and media strategies, and synthesizing a more complete understanding of the consumer. The tech’s proliferation makes it critical for business leaders to understand the capabilities of AI in order to invest their resources wisely, rather than throw money at fluff.
What does this all mean for today and the future? As entertainment marketing and AI experts, we’re here to separate fact from fiction.
First, the two essential ingredients in AI is data and computing power. A vast amount of consumer, video, or behavioral data paired with computing power can produce amazing insight. AI algorithms thrive in these types of environments and is a large reason for the recent successes in the marketing field.
Second, it is estimated that 80%-90% of data being created is unstructured. Unstructured data is data that doesn’t easily fit in a pre-defined schema. For example, images, text data, video, and audio would all be considered unstructured data. Some of the most exciting developments in AI have happened when AI algorithms are paired with vast amounts of unstructured data.
Deep learning algorithms, which can analyze both structured and unstructured data, can be very powerful. For example, Amazon’s voice recognition capabilities with Alexa are powered by a lot of data and deep learning algorithms. Training on thousands of hours of speech, Alexa interprets the unstructured data of your voice, combines it with the structured data like location, time, and date, to deliver meaningful responses to your requests.
This gets even more exciting when you have a lot of unstructured and structured data residing in one place. For example, YouTube receives about 500 hours of uploaded content every minute. A YouTube video contains both structured data such as the number of views, likes, and comments as well as unstructured data like videos, titles, and thumbnails. Google’s algorithm analyzes all of this data to make a collective recommendation for your next video in real time.
In our specific field of matching brands to the best content opportunities, deep learning algorithms are an essential component to drive value for our brand clients. To understand which influencer would be a perfect match for a brand, you would not only care about the content (unstructured data), but also the subscriber base (structured). Analyzing both of these streams of data allows us to make the most accurate predictions as to which influencers will work best for a particular brand.
This all sounds great – but how does it actually work? To simplify things, we will consider supervised algorithms. Supervised algorithms are computer programs which can take in a set of data points and learn to predict an outcome, where both the data and outcome to predict are provided.
For example, imagine you are in the market for an influencer campaign and would like to predict the value (number of conversions) of influencers with whom you have never worked. You might go to Instagram and look around at some influencers you have worked with and start to notice trends: some influencers have more followers, some get more engagement on their posts, and others are good at creating beautiful images with engaging captions. These different factors seem to be correlated with the conversions in varying degrees. You could take that topline information and attempt to apply it to other influencers to estimate how many conversions they might get. This is a great start, but humans have limited ability and time to ingest data. One can only spend so much time looking through hundreds of profiles to pick out influencer trends.
What if instead, you collect data on many influencers and the number of conversions they generated. If you could, you’d love to get data such as the number of followers they have, the images from their posts, and the number of likes and comments. You’ll notice in this example we have both structured (number of followers) and unstructured data (images) as inputs to our algorithm. Using AI algorithms, you could feed all these data points to the computer program and ask it to find patterns to predict a desired output value, in this case number of conversions.
There are a lot of benefits to letting a computer handle these prediction over a human.
- AI can quickly analyze data from many influencers at once
- It can efficiently evaluate tons of variables about any given influencer. Details like length of the video, when they are posting, use of text, clear CTAs, demographics, audience sizes, and engagements. The computer sorts out what is important to your output value.
- Once the AI learns how to effectively predict influencer conversions, if you have a large enough sample set, the algorithm can replicate those predictions on future influencers for which you don’t know the value.
- Finally, since the AI distills nuanced data inputs, one can customize it to fit an exact need, depending on the KPI your looking for whether it be reach, engagement, or conversions.
So where are the humans in this process? AI shouldn’t be seen as a replacement for an incredible execution team. Rather it takes the minutia out of the process so campaign execution teams can spend less time crunching numbers and more time working directly with promising influencers. If the AI predictive models identifies the best influencer for your needs, you can then focus your attention on outreach, influencer creative, and the ultimate deliverables. This collective AI + human approach combines the ideal skillsets from each to deliver better campaigns.
AI is always improving with additional campaign data. It continues to finetune its predictions to fit a specific brand and audience and can even help deliver insights into what content is having an impact. This cycle of data means that brands which are leveraging AI are continually getting better, more efficient, and smarter.
The AI augmented future is an exciting one. With such a massive amount of content out there, we are committed to pushing the state-of-the-art in AI to enable brands to integrate into all forms of entertainment as effectively as possible. Deep learning artificial intelligence will better understand this vast content landscape, and ultimately provide brands better ways to reach consumers through the power of entertainment.
Tyler Folkman is the Head of AI at Branded Entertainment Network
Sources:
- https://courses.cs.washington.edu/courses/csep590/06au/projects/history-ai.pdf
- https://www.forbes.com/sites/bernardmarr/2016/03/22/a-short-history-of-deep-learning-everyone-should-read/#2b0e3c6f5561
- https://www.zdnet.com/article/google-io-from-ai-first-to-ai-working-for-everyone/
- https://www.datamation.com/big-data/structured-vs-unstructured-data.html
- https://www.tubefilter.com/2019/05/07/number-hours-video-uploaded-to-youtube-per-minute/
