Back to Insights

Machine learning vs. artificial intelligence in media

November 14, 2018 By Audrey Kuah

Confused about both terms? Unsure what’s the difference between both? You’re not alone. Machine Learning (ML) and Artificial Intelligence (AI) are both often written about, but rarely understood. Unfortunately, because both Machine Learning (ML) and Artificial Intelligence (AI) are in continual development, and both revolve around computers becoming more autonomous, AI and ML’s terminology often overlap and interchange. So, what are they exactly? And what does this mean for media planners?

The difference between Machine Learning and AI

Artificial Intelligence is the creation of computer systems to be able to perform tasks that would otherwise require human intelligence. There are two ways to think about AI; narrow and general. The difference is narrow AI controls a specific task, such as operating a driverless vehicle. But general AI operates holistically around all aspects of life. General AI becomes the convergence point when a computer can think, perceive, and even feel as we do.

Machine Learning is the capability of a computer to process data and adapt that process to streamline performance. If a computer is essentially just something that processes calculations, ML enables computers to resolve these processes quickly. ML can create short-cuts to save time, and even pre-empt a course of action without having to repeat the same calculations. This is called “unsupervised learning”, where the machine optimises itself without being given a command by a user. It has mass appeal because of its precision in working out the probability of events, based on data amassed in the system. Unsupervised learning requires extensive amount of well-labelled data for the task we set the machine to optimise.

Something that links AI and ML is a process called Deep Learning. In this process, computers operate with ML capabilities, such as optimising data and operating under “unsupervised learning,” but processes much more data using thousands of processing cores. If we think of AI as a computer version of a human brain, there needs to be a computer equivalent to the billions of neurons that make up the brain. These are the processors. The more processors you have, the more powerful the ‘computer brain’. Deep Learning is a way to increase the ‘brain power’ of Machine Learning.


How does this apply to Media Planning?

Machine Learning's value is its ability to manage and extract value from the volume of data, a key feature of the digital economy. For now, ML is more important and more practical for brands. And hence infinitely more valuable for the following reasons:

Machine Learning has more industry applications, already having tried and tested statistical techniques. A key strength of ML is incorporating shortcuts (“known as heuristics”) to reduce the drain on computing power and give results faster. Like Deep Learning, ML still needs data upfront.

ML has been used to put seven statistical models on top of client data, in order to optimise media investments. A six-month test for a travel client saw a 10% increase in ROI, as the ML algorithm was programmed to optimise channel placement and spend. ML plugged the gap between imperfect data to maximise conversions. ML enables the construction of algorithms that can hunt down what data is important and then proactively target.

This matters for brands because as the digital economy grows, so does the amount of data people produce. This can lead to more touch points for brands, but ML can be used to make this relationship less complicated, and ultimately more valuable for both audience and client.

Thinking of the points below can help brands and marketers to get started with using ML:

  1. Begin by asking, what can be achieved to ensure we are capitalizing on AI solutions in the paid media space?
  2. Establish an objective you can measure afterwards. This objective must be easily relatable to ML solution and what data you can use.
  3. Understanding the data you use will determine the successful of a ML campaign:
    • Acquire the most granular and detailed data, ingesting it as frequently as possible.
    • Ensure the data qualifies as accurate to the campaign.
    • Apply a consistent data structure.
    • Quantity linked to metrics used to maximize paid media effectiveness.
  4. Recognise the fact that ML solutions take time to bring to fruition and if there is a realistic timeframe to utilise this technology.
  5. Use a table below to understand the potential problems you may face so you can mitigate against them.

Can AI be applied to media planning?

AI is still a long way from being a mainstream product. This is because there is an evolutionary process, starting with ML to Deep Learning and eventually to AI controlling a specific task then to AI operating holistically around all aspects of life

Standard computers don’t think for themselves, and the way they operate is hard-coded into the computer’s operating system. Creating Artificial Intelligence – where a computer is aware of what it is doing and is entirely autonomous – is a huge leap from where the industry is right now. However, we are developing computers that can process information quickly via machine learning. Based on advanced algorithms, ML can associate that information with other data to deliver results faster and more accurately. It’s only through ML that we have the potential, perhaps, to reach AI.