ad:tech London - Exploring the AI promise in AdTech

LinkedInTwitter

GeoSpock ad:tech London workshop image

GeoSpock hosted a workshop entitled ‘Exploring the AI promise in AdTech’ at the recent ad:tech London 2017 show. It was a standing room only workshop, with a mix of presentations and a panel discussion. Panellists included:

GeoSpock Founder and Chief Technology Officer Steve Marsh kicked off the session with a quick introduction.

“AI creates winners – it is no coincidence that the likes of Google, Amazon and Facebook are big investors in AI,” Marsh said. “AI is a system that attempts to maximise a certain set of outcomes. Machine learning is one of the techniques used to create AI.

“We often talk about deep learning as well. Machine learning has to define outcomes ahead of time, but deep learning pulls out those patterns for you.”

Here are the key takeaways from the workshop:

1) Be definitive

Confusion reigns across AdTech and various other sectors when it comes to knowing the differences between AI and machine learning – not to mention deep learning and analytics. The members of the panel agreed that the definitions need to be nailed down so that misunderstandings are ironed out.

“The terms are bandied around, blurring the lines and making it more difficult for people to understand what’s going on,” Michael Nutley, Journalist, Editor and Content Strategist, said. “People think that robots are going to kill us or take our jobs and that’s not a helpful basis for discussion, but it seems to be the overriding theme. It’s hardly surprising no one really knows what’s going on.”

Paul Gubbins, a Programmatic Columnist and Consultant, said: “AI is used as an umbrella term and that’s causing quite a bit of confusion. Machine learning and AI will be there to help us do our jobs better and make us capable of doing things we aren’t doing today.”

The panel also agreed that some companies might be getting away with falsely using the term ‘AI’ to sell their services, taking advantage of a client’s lack of understanding.

According to Blippar’s Global Head of Experience Strategy, Omaid Hiwaizi, AI is “technically the entire field”, with deep learning, machine learning and analytics sitting underneath. “Deep learning is a computer system ingesting huge amounts of data and clustering that information into relevant concepts and ideas. Machine learning is when a system has already started to understand criteria about a certain dataset and a human has set some of the criteria by which it can be judged.”

 

 

2) Application

A variety of different AI techniques are being utilised to solve the “data problem”, according to Marsh.

Machine learning is in its early stages of application, but it is being used, for example, in the fields of campaign performance and fraud identification.

“AI will come into its own from an advertising perspective when it begins looking at different creatives and starts to build creatives,” Gubbins said.

Marsh added: “Current machine learning techniques reach a limit, at which you have to turn over to deep learning, which can improve understanding of the data. Deep learning is brute force in many ways, but machine learning is more tailored.

“Today we’re talking about desktop, mobile and tablets, but in 24 months almost every device in the home will be connected and every agency and advertiser will want those outputs. That is where AI and machine learning will start to get massive traction. Trillions of data points will be multiplied by the Internet of Things.”

3) Handling the data

Managing such an awesome volume of data will be an immense challenge in the future, the panel agreed.

“Even today, we’re processing trillions of rows of data and there’s no way a person can look at all of that,” Marsh said. “So, you need machine learning and then deep learning techniques.

“It will be important to extract insights on a very large scale and I don’t see a future where AI isn’t deployed everywhere.

“Oftentimes people are gathering large amounts of data but it takes too long and too much money to process it.

“At GeoSpock, we can handle that vast amount of data and can do it more cost-effectively, getting insights in real time. This will allow you to carry out self-optimising campaigns, putting into context what every user is doing. It’s very powerful.”

4) The ‘black box’ challenge

Trust is a big obstacle in AI, with the workshop’s moderator, Campaign’s Global Technology Editor, Emily Tan, asking how clients can be expected to trust techniques that they do not fully understand.

There was a consensus across the board that transparency will be key in the future, with Tan explaining how there have been calls for a moral code within the field of AI.

“If you think it’s a black box, DSP (demand-side platform) vendors should be clear with you and lift the hood to show you how it works,” Gubbins said. “People have to work with transparent partners.”

Nutley added that “a distinction needs to be drawn”. He explained: “You might not necessarily be able to say what the system is doing right now, but you can see what the parameters are. By looking at the outputs, you can draw conclusions about what is going on inside, and the methodology can be explained.”

GeoSpock’s efficient data indexing and management capabilities will be essential for the future of ad tech. For more information about how GeoSpock can help ad tech firms to engage more effectively with consumers, click here.