In the #PlastreamCoffee cycle we will be talking to the people standing behind the technology at Plastream and will bring closer technical and business issues related to our tool. In our first interview, we will be talking to Łukasz Czarnecki, the AI architect.
If you could describe the operation of Plastream in two sentences, how would you do it?
Plastream uses artificial neural networks (Deep Learning) for automatic picture analysis: both static and videos. Because we automatically receive information on what appears we can match it with e-commerce products. Thanks to this we can successfully support people in context matching ads to visual content.
Will artificial intelligence replace humans or will it support their work? Is it a threat or helpful for business?
Not many people remember that at the end of the XIX century Thomas Edison organised public shows where he shocked animals with alternating current to show people how dangerous it is to humans. Today alternating current is a normality that we don’t even think about. I assume it’s the same with AI which stimulates imagination in the direction of the vision from the “Terminator”. In reality, we already see many benefits, for example in medicine where diseases are easily detected, or new treatments are discovered. We can’t foresee the future of AI, but from a business perspective, it’s an open door for new chances and opportunities.
Let’s talk about publishers. How can AI help in everyday work for publishers from a technical side?
We can think about it this way that if a particular task is being done by a person but doesn’t require in-depth analysis or broader context knowledge (history, culture, society, etc.), it can be automated. Example: we can automatically define a happy Donald Trump in the picture, but can’t assume why the montage on the cover is controversial. We won’t automatically generate a funny text about the most recent sitting of the Sejm, but we can automatically define that it’s a text about Mateusz Morawiecki, the topic is about tax, and the content is either positive or negative. Of course, these technologies are developed continuously so more advanced tasks will be automated.
E-commerce fights for clients every day. How can AI support these actions?
During the first step, these actions will have to be supported by AI or at least by an analysis of data. The battle over consumers is tight, and aside from creating needs you must focus on existing needs and come to a potential client with a ready answer. Each missed message is a missed chance. On the other hand the awareness of the user in using their data increases and will continue to increase. Advertisements must be matched to the user, but the user doesn’t want to share much about themselves. The key is the information about the context in which we have a chance to communicate. But how do you automatically get to know the context of specific internet content (texts, pictures, videos)? This is where AI comes to help.
Is polish business ready for AI? How does this situation look?
Every case is different. Every company before starting with AI should gain awareness about their data. This connects not only with the knowledge of what kind of data is processed but also the way it’s collected. A company that uses this conscious approach-collects as much data as possible and enables availability may consider AI projects. While a company with great potential that doesn’t use this approach has some work ahead of them. The earlier, the better. In a reality where data has a more significant worth, it’s easy to wake up with remorse “Why didn’t we start collecting data 2 years ago?”.
Similarity Search is one of the “driving motors” of Plastream. How does it work from the technical side?
The key here is the similarity of objects that don’t connect with the similarity of the whole picture. For example, a picture of Doda in a red dress is different from the picture of this dress on the website. It’s still the same dress, and we want to and can automatically capture the similarity. The first step is finding the dress in the picture and the second is capturing the features (colour, material, etc.). Because we already have an analysed e-commerce product database we can compare them to Doda’s dress and chose the most similar one.
Plastream has carried out tests. Can you tell us more about them?
Part of the tests that we are carrying out has a purely technical character. Along with our clients, we try to check if our technology works on their data as we assume. The second type of tests is checking in what way technological worth that we bring works on business. I can only say that we work with big players on pioneering AI applications in not only media. Most industries will change under the influence of the development of AI, and we want to be a part of this change.