Uber is, so far, the most popular ride-sharing application, with over 93 million users globally. The application was a revolutionary solution back in 2009; for over 13 years, it has kept increasing revenue. In 2015 Uber was one of the first to adopt machine learning for processing big data. Within the next 3 years, the application reached AI at scale, becoming one of the handfuls of companies that have successfully integrated ML strategy within a short period.
Let’s dig deeper to understand Uber’s value proposition and the success formula.
How Important Is AI in Uber-Like Apps?
The basic notion of Uber-like apps is to build a highly scalable platform for drivers and customers. Riders are looking for a modern solution that is fast, reliable, personalized, and cost-effective; drivers need to optimize time earnings while at the same time maintaining the quality of work. With advancements in AI, all the features that once seemed unrealistic are already actively in use. And here is how AI is changing the ridesharing experience.
- Predicting and anticipating demand patterns to design optimal supply for each market’s mobility demands.
- Predicting and improving ETAs (Estimated Time of Arrivals) and ETD (Estimated Time to Destination) with high accuracy.
- Getting deep insights into customer behavior and preferences to later generate a more personalized approach.
- Enhancing safety by continuously learning from patterns and listening in real-time to events.
Of course, the role of AI is much more prominent in Uber-like apps, and the technology will continue changing the industry.
Uber AI in 2022
In 2009 AI was only an idea, but Uber was already built with artificial intelligence at its core. Today, the app uses AI in most of its services; customer support prioritization, ETA (estimated time of arrival) and destination predictions, traffic forecasting, and more.
Aside from its product offerings and new enhancements with AI and ML, Uber added enormous value to the corporate AI community.
On the technological side, Uber built a machine learning platform infrastructure Michelangelo, a proprietary platform covering the core machine learning process phases. The objective was to facilitate building and deploying machine learning solutions at scale. While most businesses deploy machine learning models on one-off use case basis and must custom construct or outsource supporting software infrastructure, Uber has structured the entire process using Michelangelo. The Michelangelo platform addresses:
- Data Collection and Preparation Process – connects to multiple data lakes, pushes that data through custom pipelines to key features, and saves those features in data feature stores;
- Model Training – it utilizes the characteristics to train a variety of different machine learning models and measure their success;
- Model Evaluation – it generates reports per model type on important performance criteria and compares them across models to identify the best one;
- Online Deployment and prediction based on data entered by a consumer, for example, when they order an Uber.
The core of all these processes is Uber AI, the center for AI research and platforms that work on empowering its own applications with NLP (natural language processing), computer vision, deep learning, sensor processing, and so much more. Significantly contributing to AI community development, Uber helps other community members benefit from its platforms through open-source projects and publications. It means it is now possible to develop as advanced and AI-powered on-demand app As Uber.
AI in Uber: Advancing Mobility With Artificial Intelligence
Even though Uber has already become groundbreaking in implementing AI, this year (2022), Uber AI research has resulted in considerable advances in demand prediction and more smooth pick-up experiences. Let’s point out the most significant achievements.
Leveraging Computer Vision to Make Riding Safer and More Comfortable
It is clear now that Uber takes everything seriously when it comes to AI. One of the ways to leverage the technologies is the Computer Vision Platform which works on validating driver identity when going online. The automated deep learning transcription technology allows Uber to keep under control the growing number of drivers and ensures customers’ safety.
Enhancing Real-Time Forecasting With Neural Networks
Real-time operation is at the core of Uber and Uber-like apps. Tracking performances online has become more effective by utilizing ML models powered by neural networks that help forecast pick-up and drop-off ETA, rider demand, and determine hardware capacity requirements. Uber AI developed three more tools to reach KPIs and improve forecasting: X-Ray, Gennie, and HotStarts, each responsible for processing big data and improving predictions.
Creating More Seamless Communication With Conversational AI
Customer service is the major function of Uber that is treated with the most excellent care. To support top-notch end-to-end experience and at the same time provide easy access Uber has adopted conversational AI technology to help support personnel answer user concerns correctly and quickly. Furthermore, the platform managed to reduce distracted driving by allowing drivers to connect with customers more effortlessly through hands-free pick-up and one-click chat.
Intended to go well beyond what has already been done, Uber launched an open-source Plato Research Dialogue System, a conversational AI platform for developing, training, and deploying conversational AI agents, allowing cutting-edge conversational AI research.
Improving Location Accuracy With Sensing and Perception
AI and ML technologies were implemented in almost all core functionalities of the ridesharing app, including vehicle location. To facilitate the processes, Uber launched the Sensing and Perception platform that worked on improving the coverage, speed, and accuracy of GPS services.
Overcoming GPS constraints and having more exact locations allows riders and drivers to locate each other more easily, minimizes cancellations on both sides, and improves estimated arrival time (ETA).
Publishing Original AI Research
Uber has its greatest contribution to AI integration in all industries, not limited to on-demand services. As a valuable asset, Uber released research publications on its projects, topics on conversational AI and neural networks, blog articles, and open source codes – EvoGrad Python library for prototyping natural evolution-like algorithms for training ML models, and Hypothesis GU Funcs (a Python package for unit testing). All these and many more resources that Uber shared contribute to AI development and the ability to incorporate similar technologies into other apps.
Fostering AI Collaboration Through Open Source
Committed to sharing as much information and knowledge with the community as possible through open source objects, Uber AI lab released Ludwig’s open-source deep learning toolbox for users to train and test deep learning models without coding. The toolbox has also been adopted by IBM, Nvidia, and Apple.
Before on-demand apps like Uber, we had a different ridesharing culture with its benefits and drawbacks. Today we enjoy all the comfort of ridesharing apps looking forward to innovations.
One of the most significant and ground-changing innovations we will have by 2025 is the Green Future program, with thousands of drivers switching to electric vehicles (EVs). The program will also contribute to the development of AI and self-driving cars.
Promoting a zero-emission mobility platform, by 2040, Uber promises to serve 100% of rides globally with zero-emission vehicles. For sure, there will be new ways to ride green and provide sustainable alternatives for transportation.