Tech giants DiDi, Baidu and Alibaba will be adopting NVIDIA’s artificial intelligence software.
NVIDIA’s new Pegasus chip promises to deliver self-driving taxies by 2018. TechRepublic’s Alison DeNisco details the new platform, and why self-driving cars could arrive sooner than expected.
During his keynote address during the GPU Technology Conference in Suzhou, China, NVIDIA founder and CEO Jensen Huang announced a bevy of new software, hardware and partnerships with some of the biggest technology companies in China.
Huang said the world’s largest e-commerce company, Alibaba, was already using NVIDIA’s artificial intelligence platform to provide users with recommendations and the search engine Baidu would be doing the same soon.
For Alibaba, the partnership has already been fruitful. Nov. 11, known as Singles’ Day in China, has become as massive shopping event and generated a reported $38 billion in sales this year, double the amount seen on Black Friday and Cyber Monday combined.
Alibaba’s director of heterogeneous computing, Lingjie Xu, said: “We deploy state-of-the-art AI technology at massive scale using the NVIDIA accelerated computing platform. The platform’s intuitive search capabilities and reliable recommendations allow us to support a model six-times more complex than in the past, which has driven a 10% improvement in clickthrough rate. Our largest model shows a hundred times higher throughput with T4 compared to CPU.”
During his speech, Huang also unveiled the seventh generation of the company’s TensorRT inference optimization software, which will enable real time, end-to-end conversational AI. NVIDIA is also making huge inroads in providing systems for cars, announcing another partnership with the Chinese rideshare company DiDi.
“We have entered a new chapter in AI, where machines are capable of understanding human language in real time,” said Huang told the audience.
“TensorRT 7 helps make this possible, providing developers everywhere with the tools to build and deploy faster, smarter conversational AI services that allow more natural human-to-AI interaction.”
SEE: Artificial intelligence: A business leader’s guide (free PDF) (TechRepublic Premium)
Paresh Kharya, NVIDIA’s director of product management for accelerated computing, and Danny Shapiro, senior director of automotive for NVIDIA, explained the technology further in interviews with reporters.
Providing recommendations for a huge e-commerce site like Alibaba or a search engine like Baidu is not easy and required years of work due to the massive amount of computing power needed to handle the billions of options available.
There are billions of products available on these websites and trillions of webpages to search to provide users with what they prefer.
“In order to recommend the right product to the right user at the right time, you need to model the users’ preferences and take in a lot of different variables to model user preferences,” Kharya said. “Similarly, you need to model every single item that’s out there. Whether it’s a restaurant, a news article, or whether it’s a product that’s available for online shopping.”
“The intersection of these two, the items and their parameters, users and their preferences, when you combine these two things together, you’re basically looking at a table that has billions of dimensions. In the case of Baidu, they have their data set that has 100 billion dimensions in order to provide effective recommendations.”
The other issue companies like Baidu face are the constantly changing preferences of users. Kharya explained that Baidu wants their recommender models to be updated every few minutes to match changing user catalogs, meaning the system needs to be retrained and fine-tuned constantly.
The same goes for two of Alibaba’s biggest e-commerce platforms, Taobao and Tmall, the executives said. The scale of the operations has grown so immense that they needed new, innovative solutions to deal with the deluge of activity they deal with, especially on big shopping days like Singles’ Day.
One of the biggest challenges Alibaba faces is the sheer number and volume of products they have in their catalog, which is now well over two billion products. The company also has to cater to 500 million users all shopping at the same time on a holiday like Singles’ Day.
“In order to match these two billion products to the user preferences is really a big task. Two billion products means that if a single user is browsing for these products in a catalog, taking just one second to look at one product would take 32 years just to browse through all these products,” Kharya said.
“Recommendations are vital in order to expose the right product to the right user at the right time. Alibaba uses advanced recommender systems running on NVIDIA’s T4 GPUs to recommend products. These products are shown when users log into their Taobao and Tmall e-commerce apps. When they’re adding products to their shopping carts, the related products are shown, or when they are selecting products to try to find more information about what other products that can go together with these products.”
Shapiro explained that DiDi, which he called the world’s largest mobility and transport platform, chose NVIDIA for its autonomous vehicles and cloud infrastructure.
DiDi is using NVIDIA AI for training through testing and validation for in-vehicle AI, helping the company fuse all the sensors running and allowing it to make safe driving decisions. Their system will help vehicles detect objects like cars, pedestrians and bikes as well as tasks like lane detection, sign detection and parking.
Other innovative tools, like NVIDIA DRIVE, will help autonomous vehicles sort and make sense of the deluge of data constantly being ingested by different sensors, Shapiro said.
“The AI autonomous vehicle is a software-defined vehicle required to operate around the world on a wide variety of datasets,” said Huang.
“By providing AV developers access to our DNNs and the advanced learning tools to optimize them for multiple datasets, we’re enabling shared learning across companies and countries, while maintaining data ownership and privacy. Ultimately, we are accelerating the reality of global autonomous vehicles,” Huang said.
“Creating a safe autonomous vehicle is perhaps society’s greatest computing challenge.”