[ad_1]
When Gao Hongzhi was young, he lived with his family in Gansu, which is located on the edge of the Tengger Desert in central northern China. Recalling his childhood, he recalled the continuous earth wind outside their house. For most of the year, less than a minute after going out, the sand would fill up any empty space and crawl into his house. inside. Pockets, boots and his mouth. The monotony of the desert lingered in his mind for many years. During his college years, he transformed this memory into the idea of building a machine that could bring plant life to the desert landscape.
Efforts to stop desertification (the process by which fertile land turns into desert) have focused on expensive artificial solutions. Hong Zhi designed a robot that uses deep learning technology to automate the tree planting process: from determining the best location to planting tree seedlings to watering. Although he has no experience in artificial intelligence, as an undergraduate, Hong Zhi used Baidu’s deep learning platform PaddlePaddle to splice different modules together to build a robot with better target detection capabilities than similar machines already on the market. Hong Zhi and his friends completed the development of the final product and put it into use in less than a year.
Hong Zhi’s desert robot is a vivid example of the increasing popularity of artificial intelligence.
Today, more than 4 million developers are using Baidu’s open source AI technology to build solutions that can improve the lives of people in the community, many of whom have little technical expertise in this field. “In the next ten years, artificial intelligence will become the source of changes in the structure of our society, changing the way industries and companies operate. This technology will expand the human experience by allowing us to explore the digital world more deeply,” Baidu CEO Robin Li Presented at Baidu Create 2021, an artificial intelligence developer conference.
As we entered a new chapter in the development of artificial intelligence, Baidu’s chief technology officer Wang Haifeng identified two key trends that will support the industry’s path forward: artificial intelligence will continue to mature and increase its technological complexity. At the same time, deployment costs and barriers to entry will be lowered-this is beneficial to companies building artificial intelligence solutions on a large scale and software developers exploring the world of artificial intelligence.
Fusion of knowledge and data with deep learning
The integration of knowledge and data with deep learning has significantly improved the efficiency and accuracy of artificial intelligence models. Since 2011, Baidu’s AI infrastructure has been acquiring new information and integrating it into a large-scale knowledge graph. Currently, this knowledge graph has more than 550 billion facts, covering all aspects of daily life, as well as topics in specific industries, including manufacturing, pharmaceuticals, law, financial services, technology, media, and entertainment.
This knowledge graph and massive data points together form the building block of Baidu’s newly released pre-training language model PCL-BAIDU Wenxin (ERINIE 3.0 Titan Edition). This model is superior to other language models without knowledge graphs on 60 natural language processing (NLP) tasks, including reading comprehension, text classification, and semantic similarity.
Cross-modal learning
Cross-modal learning is a new field of artificial intelligence research, which aims to improve the cognitive understanding of machines and better imitate human adaptive behaviors. Examples of research work in this area include automatic text-to-image synthesis, in which models are trained to generate images from text descriptions only, and algorithms for understanding visual content and expressing comprehension in words. The challenge of these tasks is to allow the machine to establish semantic connections between different types of data sets (such as images, text) and to understand the interdependence between them.
The next step of artificial intelligence is to integrate artificial intelligence technologies such as computer vision, speech recognition, and natural language processing to create multi-modal systems.
In this regard, Baidu has introduced a variant of its NLP model that links language and visual semantic understanding. Examples of practical applications of such models include digital avatars that can perceive the surrounding environment like humans and handle customer support for companies, and algorithms that can “paint” artworks and create poetry based on their understanding of the generated artworks.
This technology has more creative and influential potential results. The PaddlePaddle platform can establish cross-visual and linguistic semantic connections, which allowed a group of Chinese master students to create a dictionary to protect endangered languages in areas such as Yunnan and Guangxi by making it easier to translate them into simplified Chinese.
AI integration across software and hardware and specific industry use cases
As artificial intelligence systems are applied to solve increasingly complex and industry-specific problems, more emphasis is placed on optimizing software (deep learning framework) and hardware (artificial intelligence chips) as a whole, rather than optimizing them individually, taking into account the following factors: such as computing power , Power consumption and latency.
In addition, the platform layer of Baidu’s artificial intelligence infrastructure is undergoing tremendous innovation, and third-party developers are using deep learning capabilities to build new applications tailored to specific use cases. The PaddlePaddle platform has a series of APIs to support the application of artificial intelligence in new technologies such as quantum computing, life sciences, computational fluid dynamics, and molecular dynamics.
Artificial intelligence also has practical uses. For example, in Shouguang, a small city in Shandong Province, artificial intelligence is being used to streamline the fruit and vegetable industry. It only requires two people and one application to manage dozens of vegetable sheds.
It is worth noting that Wang said, “Despite the increasing complexity of artificial intelligence technology, the open source deep learning platform integrates processors and applications like an operating system, reducing the number of companies and individuals who want to integrate artificial intelligence into their businesses. Barriers to entry.”
Lower barriers to entry for developers and end users
In terms of technology, the pre-training of large models such as PCL-Baidu Wenxin (ERNIE 3.0 Titan Edition) has solved many common bottlenecks faced by traditional models. For example, these general models lay the foundation for running different types of downstream NLP tasks (such as text classification and question answering) in a unified place. In the past, each type of task had to be solved through a separate model.
PaddlePaddle also has a series of developer-friendly tools, such as model compression technology, used to adjust general models to suit more specific use cases. The platform provides an officially supported industrial-grade model library, with more than 400 models, from large to small, only a small part of the general models are retained, but comparable performance can be achieved, and model development and deployment costs can be reduced.
Today, Baidu’s open source deep learning technology supports a community of more than 4 million AI developers, who have created 476,000 models and contributed to the AI-driven transformation of 157,000 companies and institutions. The examples listed above are the result of innovations in various layers of Baidu’s artificial intelligence infrastructure. It integrates technologies such as speech recognition, computer vision, augmented reality/virtual reality, knowledge graphs, and pre-training large-scale models, which is one step closer to perception. The world is the same as human beings.
In the current state, artificial intelligence has reached the maturity level that can accomplish amazing tasks. For example, without PaddlePaddle’s platform, the recently launched Metaverse XiRang would not be able to create digital avatars for participants from all over the world to connect through their devices. In addition, future breakthroughs in quantum computing and other fields can significantly improve the performance of the meta-universe. This shows how Baidu’s different products are intertwined and interdependent.
In a few years, artificial intelligence will be close to the core of our human experience. For our society, what steam power, electricity and the Internet will be to previous generations. As artificial intelligence becomes more and more complex, developers like Hong Zhi will work more as artists and designers because they are free to explore use cases that were previously thought to be only theoretically possible. The sky is the limit.
This content is produced by Baidu. It was not written by the editors of MIT Technology Review.
[ad_2]
Source link
