Artificial Intelligence has been a game changer for the business world, as new breakthroughs allow companies to unlock more and more value from their data. One such breakthrough, deep learning, has revolutionized AI across several domains, ranging from computer vision and natural language understanding, to machine translation and speech recognition.
Deep learning uses large neural networks with multiple layers of processing units, taking advantage of the advances in computing power and improved training techniques to learn complex patterns across billions of data points.
While deep learning has numerous business applications, classifieds and online marketplaces in particular, have been paying close attention. Companies such as eBay, Carousell and Letgo are leveraging insights from deep learning models to reinvent the age-old experience of people buying and selling online. These companies, which have billions of commerce data points and transaction data, are turning to machine learning models to derive greater insights, from recognizing objects in listings, to personalizing recommendations and even identifying the buying intent of users.
eBay, for example, launched its image search feature that allows users to find what they want by simply taking a photo. This was done through training a deep learning model called a convolutional neural network that scans and matches images across eBay’s database. Online marketplace, Carousell also utilizes a similar deep learning model to help its users list items faster. When a user creates a listing, the Carousell app scans photos taken by users during the selling process and suggests categories and product titles.
The impact of AI to drive key business results is real. As one of the pioneers in the field, eBay reported that AI-powered improvements drive more than $1 billion per quarter in sales on its marketplace. With that kind of business impact, AI plays such a crucial role in the present and future of classifieds and online marketplaces. Companies need to rethink, even going to so far as to remake themselves in order to succeed in an AI-first world. But, what does an AI-first organization mean in practice?
In my experience, AI-first organizations need both centralized and decentralized AI functions. Centralized AI research teams are tasked with keeping up with technology trends, whether they create new research projects or even develop new applications. Successful research can then be put into production, by working together with engineering teams.
While a centralized AI function is a necessary first step for any organization, decentralizing AI know-how to product teams is probably the key piece to enabling organizations to ship AI features for business impact. Educating internal designers and product managers, ensures these features are designed and experimented frequently. Disseminating this knowledge can be achieved by embedding data scientists and machine learning engineers in the various product teams to work with product managers and software engineers.
Besides education, internal training for engineers allows them to familiarise themselves and try out new technology. Major tech companies are embracing programmes such as Facebook AI Academy, where engineers are offered training opportunities in deep learning classes or immersion program inside AI research teams.
The importance of AI and deep learning can only increase for all companies, not just for classifieds and online marketplaces. As the volume of unstructured image, video, text and voice data increases, there is an opportunity for companies to use AI and derive valuable insights for their customers. Therefore, building up one’s AI capabilities, and positioning organizations to be AI-first will be critical for success.