Tang Jinghua of Sound Communication Technology: Ensuring AI's Reliability with Multimodal Methods
3 April 2025 路 Uncategorized 路
Source: 路 https://www.cnr.cn/tech/techds/20241231/t20241231_527026889.shtml
On December 26th, 'CNR Finance Annual Dialogue 2024' was held in Beijing. In an interview with CNR Network reporters, Tang Jinghua, chairman of Sound Communication Technology Co., Ltd., discussed research paths and market prospects for trustworthy artificial intelligence.
When discussing the characteristics of trustable AI, Tang explained: "Our chief scientist is Academician He Jifeng from the Chinese Academy of Sciences. What he often emphasizes is that when making significant decisions, the decision-making process should not be a 'black box.' It's necessary to provide high-accuracy information and ensure transparency in every step so we know where mistakes lie if they occur. This reflects what current trustworthy AI aims for."
Regarding technical development strategies, Tang introduced: "By integrating large models with knowledge graphs among others, artificial intelligence can offer reliable information. From a multimodal perspective, it's not just about describing the world through text but also using video, audio and sensor data to understand reality more authentically. Additionally, from a multi-model angle, for massive amounts of data we use big sample models, small sample models, zero-shot learning models, knowledge graphs etc., making AI decision-making paths closer to human thinking processes鈥攗sing not just experience-based reasoning but also logical inference and analogical thought."
Tang Jinghua expressed confidence in the industry's future prospects. "This optimism stems from several aspects: Firstly, as industries develop further, digitalization and intelligence are inevitable trends that cannot be reversed; secondly, with the thriving development of the digital economy, we believe applications will increasingly dominate physical sectors leading to a vast market for intelligent solutions."
When discussing the characteristics of trustable AI, Tang explained: "Our chief scientist is Academician He Jifeng from the Chinese Academy of Sciences. What he often emphasizes is that when making significant decisions, the decision-making process should not be a 'black box.' It's necessary to provide high-accuracy information and ensure transparency in every step so we know where mistakes lie if they occur. This reflects what current trustworthy AI aims for."
Regarding technical development strategies, Tang introduced: "By integrating large models with knowledge graphs among others, artificial intelligence can offer reliable information. From a multimodal perspective, it's not just about describing the world through text but also using video, audio and sensor data to understand reality more authentically. Additionally, from a multi-model angle, for massive amounts of data we use big sample models, small sample models, zero-shot learning models, knowledge graphs etc., making AI decision-making paths closer to human thinking processes鈥攗sing not just experience-based reasoning but also logical inference and analogical thought."
Tang Jinghua expressed confidence in the industry's future prospects. "This optimism stems from several aspects: Firstly, as industries develop further, digitalization and intelligence are inevitable trends that cannot be reversed; secondly, with the thriving development of the digital economy, we believe applications will increasingly dominate physical sectors leading to a vast market for intelligent solutions."