In the first of the Delivery Hero’s new technology series – “Delivery Tech Talks”, we are grilling (pun intended) Benjamin Mann, our Chief Technology Officer about Generative AI (Gen AI), automation, meaning of ‘dishing up images’, and what lies ahead for everyone’s favorite technology.
Nearly two years after Chat GPT landed on earth, everyone’s still talking about Generative AI (GenAI), but is it really more than having the right chatbot?
Gen AI has been an obvious game changer in most if not all sectors, though the concept itself is not new. Tech companies have been using it in some shape or form for years, like Amazon’s Alexa device for example. The big shift is in how newer GenAI tools like Chat GPT and Dalle E revolutionized the consumer experience of technology from better conversational customer interactions i.e no more ‘one word’ responses, to dishing up images using generative AI. These newer AI algorithms have driven changes in our own operations, for example, better prediction of efficient routes to improve customer delivery times and our own in-house employeGPT to save employees time on boring work stuff. So far, incorporating AI-enabled and automation functionalities into our systems and infrastructure is saving us EUR 140 million per year in our customer service area alone. And we are still counting…
‘Dishing up AI’ images sound like a new Netflix reality show, what does it mean?
Food is visual and ordering should be easy. Capturing appealing, accurate images, which is crucial for customers who want to see before eating, can be challenging for restaurants. By using a self-hosted Stable Diffusion Gen AI model, we can help them to quickly dish up tailored AI-generated food images that look exactly like what they have on offer, whether it’s a vegan burger or a thai curry. We also use E5, an open-source AI model, to create representations of all our dishes and restaurants in a global semantic vector space, so our Search and Recommendation Engines can find links between dishes to recommend food that customers will love but just don’t know about yet. These are just two of the ways we leverage AI and automation tools to create more unique food adventures for our customers.
Tech future is bright, but will it still be Gen AI?
We will see Gen AI change from a doer to more of an enabler, particularly for businesses – future AI models will be more about making existing AI systems better. For example, we use Meta’s open-source model Llama 3.1 to create key training data that in turn improves our own models for personalisation and discovery. Like others in the sector, we started by using readily available APIs, but now our use cases are more sophisticated and have outgrown these general purpose models. So, we have turned to using open source components or building new in house technology that can be easily tailored to our specific business needs. Another area to watch is cybersecurity, particularly fraud detection. Many AI systems can already identify suspicious patterns and transactions in real-time to reduce losses associated with fraud. But we expect to see an increase in AI powered automation of data security that would benefit everyone in the food delivery ecosystem – customers, restaurants, businesses, riders and partners – and across all sectors. Back to your first question, it is less about chatbots for us, and more about saving operational costs and improving business productivity to create amazing customer experiences.
If you like what you’ve read and you’re someone who wants to work on open, interesting projects in a caring environment, check out our full list of open roles here – from Backend to Frontend and everything in between. We’d love to have you on board for an amazing journey ahead.
