TL;DR:
- Baidu refutes claims of ties to the Chinese military’s AI training program involving Ernie Bot.
- Researchers used Baidu’s large language model (LLM) to predict combat strategies based on historical data.
- LLMs GPT3.5 and GPT4 were also employed in the training initiative.
- The study aimed to enhance AI-powered military systems’ understanding of human intentions.
- Baidu’s stocks declined in response to the news, but the company denies any connection.
- Baidu emphasizes its commitment to responsible AI usage and corporate practices.
- Ernie Bot offers various AI capabilities and has shown improved content quality.
- China’s Defense Ministry calls for responsible AI usage in military applications.
- China’s third-generation quantum computer, Origin Wukong, achieves significant milestones.
Main AI News:
Amidst swirling speculation, Baidu, the tech giant renowned for its generative artificial intelligence (GenAI) chatbot, Ernie Bot, has vehemently denied any association with a reported military training endeavor linked to the People’s Liberation Army (PLA) of China. Last Friday, media reports hinted at a university’s involvement in training the PLA’s AI system using Baidu’s expansive language model (LLM). Researchers from the Information Engineering University divulged that they fed copious data on historical conflicts, including weapon details and deployment strategies, into the LLM. The goal was to harness this data to predict combat tactics.
An academic paper, recently published in the Chinese journal Command Control & Simulation, unveiled intriguing insights into this endeavor. The research team disclosed that various LLMs, including GPT3.5 and GPT4, were deployed in conjunction with this training initiative. The LLMs were prompted to generate combat strategies and potential outcomes by drawing upon the training data at their disposal. Notably, the researchers made reference to the US military’s 2011 intervention in Libya, leveraging tactics from both sides of the conflict to fine-tune Ernie Bot’s predictive capabilities, especially concerning the US army’s strategy.
The researchers emphasized the value of these simulated results in aiding decision-making processes and enhancing the LLM’s knowledge and cognitive understanding. Their ultimate objective was to bolster AI-powered military systems, enabling them to decipher human intentions—an imperative consideration given the inherent unpredictability of human adversaries.
Baidu’s stocks witnessed a notable dip in response to the revelations surrounding this research initiative. However, the Chinese tech giant has since refuted any connections with the paper’s authors or the university in question. In an official statement addressing the academic paper, Baidu underscored that Ernie Bot remains accessible to the general public. Furthermore, it clarified that it has not engaged in any form of business collaboration or tailored services for the research team or affiliated institutions.
Baidu asserted its commitment to conducting AI-related operations in full compliance with relevant laws, regulations, and best corporate practices. It also noted that the original South China Morning Post article had been amended to reflect its statement. Baidu’s ongoing commitment to responsible AI usage and development remains at the forefront of its corporate vision.
In a bid to gather user feedback and enhance its platform, Baidu made Ernie Bot available to the general public last August. The company also unveiled plans to introduce AI-native apps encompassing reasoning, memory, generation, and understanding—four key GenAI capabilities. Notably, Ernie Bot has demonstrated the ability to generate text, images, audio, and video in response to text prompts. It can also deliver voice content in various local Chinese dialects, including Sichuan. Since its limited introduction in March, the AI model has honed its proficiency across more than 200 writing genres, with a remarkable 1.6-fold increase in content quality, according to Baidu. The addition of five plugins this month, including Baidu Search and data analytics, further underscores the company’s commitment to AI innovation.
Amidst these developments, China’s Defense Ministry issued a call for prudent and responsible AI usage in military applications. They expressed a keen awareness of security risks posed by AI and emphasized China’s opposition to leveraging AI advantages to compromise the sovereignty of other nations.
In a separate development this week, China’s third-generation quantum computer, Origin Wukong, has achieved significant milestones, processing 33,871 tasks for users worldwide since its inauguration on January 6. Users from over 60 countries have remotely accessed the quantum computer, with the United States leading in terms of frequency. This remarkable achievement underscores China’s progress in quantum computing, driven by a 72-qubit locally-made chip.
Conclusion:
Baidu’s swift denial of involvement in the Chinese military’s AI training program seeks to protect its corporate image and maintain compliance with ethical AI practices. This development underscores the importance of responsible AI usage in the military sector and highlights the continued growth of AI and quantum computing technologies in China, potentially signaling new opportunities and challenges for the global tech market.
