AI-Native Maps and Intelligent Upgrades: Baidu and OpenAI Lead Technological Innovation

Baidu is about to launch a brand new AI-native map, which will comprehensively apply the Wenxin large model to reconstruct the interaction of map products, bringing users a brand new experience. Users can wake up various functions and services in the map with a single click through natural language dialogue, significantly improving travel efficiency and decision-making capabilities. At the same time, OpenAI plans to launch a major upgrade for developers next month, aimed at reducing the cost of building software applications based on its AI model and accelerating development speed, in order to attract more companies to use its technology.

Industry Dynamics

ZTE Corporation released a text message anti-fraud system based on a large model.

ZTE Corporation launched the industry's first "Smart Governance" SMS anti-fraud system based on large models at the 2023 China Mobile Global Partners Conference. This system, built on the core of an anti-fraud large model, combines a generalized feature neural network and SNS social feature analysis into an integrated solution, governing from the root dimensions of scam SMS's intent and semantics, effectively enhancing the recognition and interception capabilities of scam SMS.

Baidu launches AI-native maps, expanding lane-level navigation coverage.

Baidu's Vice President Shang Guobin revealed that the company will soon launch a brand new AI-native map, which will comprehensively apply the Wenxin large model to reconstruct the interaction of map products. Users can activate various functions and services in the map through natural language dialogue. The new version of Baidu Map V19 will launch an "AI Guide," equipped with multi-turn natural language interaction capabilities, providing an experience similar to "real human communication." In addition, users can also customize the digital avatar of the "AI Guide."

QQ Browser launched PDF Reader Assistant

QQ Browser announces that the "PDF Reading Assistant" is open for experience testing applications, suitable for both mobile and computer. This feature is supported by Tencent's Mix Yuan large model, and its main functions include intelligent summarization, intelligent Q&A, multi-turn questioning, and original text location, among others. In the future, this feature will expand to various document formats such as Word, Excel, PPT, and web pages.

ZhiYuan Research Institute fully upgrades the Aquila large language model series

The Zhiyuan Research Institute announced that the Wudao · Tianying Aquila large language model series has been fully upgraded to Aquila2, with the addition of the Aquila2-34B model featuring 34 billion parameters. The new model shows excellent performance in reasoning, generalization, and other aspects, achieving significant results in scenarios such as intelligent agents, code generation, and literature retrieval. At the same time, the Zhiyuan Research Institute also released an open-source suite that includes the Aquila2 model series, a new version of the BGE semantic vector model, the FlagScale efficient parallel training framework, and the FlagAttention high-performance Attention operator set.

DeepMind's financial report shows a dual decline in employee costs and profits.

According to documents submitted to the UK government, Alphabet's artificial intelligence lab DeepMind reduced its employee costs by 39% in 2022, from nearly $1.2 billion in 2021 to about $731 million. During the same period, profits fell by more than 40%, from about $126 million in 2021 to approximately $74.9 million in 2022. To further improve efficiency, Alphabet merged Google and DeepMind's AI departments in April 2023 to better share knowledge and computing resources.

Australia will allow generative AI to be used in schools.

Australia has announced that starting in 2024, artificial intelligence, including ChatGPT, will be allowed for use in schools nationwide. Australia will release the "Australian Schools Framework for Generative Artificial Intelligence" to guide school staff, teachers, and students in the "safe and ethical" use of generative AI. The framework will be released in the coming weeks and promises to be reviewed every 12 months to keep up with technological updates.

OpenAI plans to launch a major upgrade

OpenAI plans to launch significant upgrades for developers next month, aimed at reducing the cost of building software applications based on its AI models and accelerating development speed. These updates include adding memory storage in development tools to utilize AI models, which is expected to cut application manufacturers' costs by up to 20 times, addressing partners' concerns about the rapidly increasing costs of using OpenAI's powerful models.

NetEase Youdao launches AI speaking private tutor

NetEase Youdao announced the launch of Hi Echo, a virtual spoken language private tutor powered by the "Ziyue" educational large model, available as an independent app and WeChat mini program. As a one-on-one spoken language tutor, Hi Echo offers round-the-clock multi-platform companionship, providing users with high-quality speaking practice anytime and anywhere.

Ernst & Young and IBM launch AI human resources solution

EY and IBM have jointly launched the human resources solution EY.ai Workforce, aimed at helping companies integrate artificial intelligence into key human resources business processes. This solution is powered by IBM watsonx Orchestrate, utilizing AI to automate HR tasks and processes, transforming traditional operational methods.

Investment and Financing Trends

Anysphere received an $8 million investment from OpenAI.

AI startup Anysphere announced that it has received $8 million in seed funding from OpenAI's startup fund, with angel investors including former GitHub CEO Nat Friedman and Dropbox co-founder Arash Ferdowsi also participating. Anysphere plans to create an AI-driven integrated development environment.