Technical giants DiDi, Baidu and Alibaba will use the artificial intelligence software from NVIDIA.
NVIDIA’s Pegasus AI chip could be autonomous vehicles
The new Pegasus chip from NVIDIA promises to deliver self-driving taxis by 2018. Alison DeNisco from TechRepublic describes the new platform and why self-driving cars could arrive earlier than expected.
During his keynote speech at the GPU Technology Conference in Suzhou, China, NVIDIA founder and CEO Jensen Huang announced a series of new software, hardware and partnerships with some of the largest technology companies in China.
More about artificial intelligence
Huang said that the world’s largest e-commerce company, Alibaba, already used the artificial intelligence platform of NVIDIA to give users recommendations and the search engine Baidu would do the same soon.
The cooperation has already been fruitful for Alibaba. November 11, known as Singles’ Day in China, has become a huge shopping event and generated $ 38 billion in sales this year, double the amount on Black Friday and Cyber Monday combined.
Lingjie Xu, director of heterogeneous computing at Alibaba, said: “We are implementing the latest AI technology on a large scale using the NVIDIA accelerated computer platform. Thanks to the intuitive search capabilities and reliable recommendations of the platform, we can support a model six times. more complex than in the past, which has improved the click rate by 10%. Our largest model has a hundred times higher throughput with T4 compared to the CPU. “
During his speech, Huang also unveiled the seventh generation of the company’s TensorRT inference optimization software, enabling real-time, end-to-end conversation AI. NVIDIA also makes a huge breakthrough in supplying systems for cars, and announces a new partnership with the Chinese driving company DiDi.
“We have entered a new chapter in AI where machines can understand human language in real time,” Huang told the audience.
“TensorRT 7 helps make this possible and provides developers everywhere with the tools to build and implement faster, smarter conversation AI services that enable a more natural human-to-AI interaction.”
SEE: Artificial intelligence: a guide for managers (free PDF) (TechRepublic Premium)
Paresh Kharya, NVIDIA’s director of product management for accelerated computing, and Danny Shapiro, senior director of automotive for NVIDIA, further explained the technology in interviews with reporters.
Making recommendations for a huge e-commerce site such as Alibaba or a search engine such as Baidu is not easy and requires years of work because of the enormous amount of computing power needed to handle the billions of options available.
There are billions of products available on these websites and trillions of web pages to search to offer users what they prefer.
“To recommend the right product to the right user at the right time, you need to model user preferences and use many different variables to model user preferences,” Kharya said. “Likewise, you have to model every item there is. Whether it’s a restaurant, a news article or a product that is 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 actually look at a table with billions of dimensions. In the case of Baidu, they have their data set with 100 billion dimensions to make effective recommendations. ”
The other issue that companies such as Baidu encounter are the constantly changing preferences of users. Kharya explained that Baidu wants their recommendation models to be updated every few minutes to match changing user catalogs, which means that the system needs to be retrained and constantly updated.
The same applies to two of Alibaba’s largest e-commerce platforms, Taobao and Tmall, the executives said. The scale of operations has grown so enormously that they needed new, innovative solutions to tackle the flood of activities that they had to deal with, especially on large shopping days such as Singles’ Day.
One of the biggest challenges Alibaba faces is the huge number and volume of products they have in their catalog, which is now over two billion products. The company also has to serve 500 million users who all shop at the same time during a holiday such as Singles’ Day.
“To adapt these two billion products to user preferences is really a big task. Two billion products means that if a single user browses these products in a catalog, it would only take 32 seconds to view one product just to go through to browse all these products, “Kharya said.
“Recommendations are vital to show the right product to the right user at the right time. Alibaba uses advanced recommendation systems that run on NVIDIA’s T4 GPUs to recommend products. These products are displayed when users log in to their e-commerce apps from Taobao and Tmall. When they add products to their shopping carts, the related products are displayed, or when they select products to try to find more information about which other products can go with these products. ”
Shapiro explained that DiDi, which he called the world’s largest mobility and transportation platform, chose NVIDIA for its autonomous vehicles and cloud infrastructure.
DiDi uses NVIDIA AI for training by testing and validating for AI in the vehicle, allowing the company to combine all running sensors and make safe driving decisions. Their system helps vehicles detect objects such as cars, pedestrians and bicycles, as well as tasks such as lane detection, sign detection and parking.
Other innovative tools, such as NVIDIA DRIVE, will help sort autonomous vehicles and understand the huge amount of data that is constantly being taken up by different sensors, Shapiro said.
“The AI stand-alone vehicle is a software-defined vehicle that is needed to operate on a wide range of data sets around the world,” Huang said.
“By giving AV developers access to our DNNs and advanced learning tools to optimize them for multiple datasets, we enable shared learning between companies and countries, while maintaining data ownership and privacy. Ultimately, we accelerate the reality of global autonomous vehicles, “said Huang.
“Creating a safe, autonomous vehicle is perhaps the biggest computer challenge in society.”
Big Data Insights newsletter
Master the basic principles of big data analysis by following these tips from experts and by reading insights about innovations in data science.
Delivered on Mondays
The Orin chip for NVIDIA DRIVE. Image: NVIDIA