From neuroscience to AI: How Shiyi Pickrell connects the dots in Expedia's data-driven environment

It may seem like a leap at first, but the more you think about the parallels between neuroscience and artificial intelligence, the more it makes sense. Artificial Neural Networks The AI ​​of the 1950s is the building block of today's AI and was modeled on how the brain works.

Still, when Shiyi Pickrell, Expedia Group’s SVP of Data and AI, set out to earn a degree in neuroscience, there’s no way she could have predicted that the degree would relate to future work in artificial intelligence and data science.

Neurons, like AI's Perceptrons, are meant to connect the dots, which is what Pickrell does on a practical level for Expedia and its technology partners.

“I started out learning how the brain works,” Pickrell said during a breakout session at the EXPLORE 24 Partner Conference in Las Vegas last month. “I worked in biotech, e-commerce, and AI in agriculture, and eventually became a machine learning scientist.

“Today, I'm teaching machine learning models to mimic what the brain can do, but at a much faster and larger scale. Think logic, planning, optimization, prediction, conversation. I'm taking machine learning and AI and applying it to travel.”

In fact, the purpose of this breakout session was for Expedia to discuss the applications of AI and machine learning with its partners on a more practical level. There's a lot of buzz and, frankly, jargon out there, so it was important to sift through the noise and talk about AI in a way that makes business sense.

It’s also worth focusing on what AI cannot do right now, rather than everything it can do.

“I think there will be a lot of things that AI can't do at first, but in the long term, I think AI will be able to do it,” Pickrell said in an interview with WiT. [get there]Having high-quality data is important for any business, but especially in the travel industry. How quickly you can process that data about traveler behavior – what they're browsing online, what they're browsing in your apps, what they're buying – is crucial.

“I don't think Gen AI or AI can create something out of thin air. But where Gen AI can play a role is in things like data discovery. The data is sitting in a lot of people's knowledge about their travels. We found that, for example, we can use Gen AI to create a chatbot that helps people ask, 'Where can I find this data?' If that person can transfer that knowledge to the bot, then you have a 24/7 assistant.”

The topic of AI assistants is particularly interesting for Expedia, which recently unveiled Romie, an AI-powered travel assistant that can be integrated into third-party apps like WhatsApp to plan trips in the background based on browsing data and real-time conversations.

The question is, is monetizing AI the next step in this trajectory? This wouldn’t be too different from, say, Google, the search engine showing sponsored results before organic results. Sponsored products on e-commerce websites are another example.

“I wouldn't jump to the conclusion that every use case for GenAI needs to be vetted from a monetization or AI perspective. The key is to help customers and travelers rediscover and fall in love with travel again,” Pickrell emphasizes.

“Today, there are so many friction points, people browse and dream about a trip before they even get on our platform or app, so we can’t be so transactional.

“I think if we can blend the discovery and dreaming phase with the deal phase in a seamless way, it will lead to a lot of fruitful outcomes. Bringing the dreaming and planning phase into the deal is exactly what Romy is helping us do.”

Experts like Pickrell have a lot of work to do: Gartner predicts that by 2026, 30% of new applications will use AI for personalization, resource efficiency, accuracy, automation and decision support, up from less than 5% today.

But Pickrell's approach to mass testing and implementation of AI is more nuanced, emphasizing quality over quantity and precision over brute force. He stresses that it's not the technology itself that matters, but the customer experience it enables, and he says the former approach is a common misconception about AI.

“Machine learning (ML) and AI are technologies, but it's really the experience that captivates customers. If you use ML and AI right, customers may not even notice it. It's so graceful and elegant,” she said.

Pickrell continued: “Some people might say, what's your strategy? Are you going to use more AI? But I think it's really about going back to basics and making sure your customers have a great experience.”

Though there's a high demand for advances in AI, Pickrell advocates for a balanced approach to creating, engineering, and adopting new AI tools: “Another caveat to 'buy vs build' is that many companies and their machine learning scientists are quick to want to build large-scale language models in-house from scratch because it's cool and they can put it on their resume. But the field is evolving rapidly.

“When you build something from scratch, you have a lot of parameters to train, which is expensive and slows down your time to market. Whereas an AI company can release a new version in just three to six months, and tweak it. If you augment it with in-house data using RAG (Retrieval Augmented Generation), you get something almost identical, but without spending a fortune.”

Still, it’s safe to assume that Expedia, which has been working with AI for years as an OTA and technology provider, is under pressure to continually and quickly churn out cooler, sleeker, and shinier AI solutions for its partners.

Finally, I asked Pickrell if the demands ever become unrealistic or unmanageable at times.

“Let me answer that another way,” Pickrell responded. “There are definitely a lot more people clamoring for it right now. I've definitely got more people knocking on my door. Even my internal team is saying they want to unlock it in a streamlined way.”

“In any scenario, there's a certain ratio of success to failure, so we want to give people a way to try things quickly without losing sight of their priorities. Internally, we provide the Gen AI platform, which allows teams to leverage different resources – open source, commercial, and even in-house solutions – so everyone can work faster.”