Out in Front

Audrey Tang, Andrew Ng share insights on Taiwan's strengths and AI regulation

Ines Lin, Taipei; Vyra Wu, DIGITIMES Asia 0

Audrey Tang, center left, and Andrew Ng, center right. Credit: DIGITIMES

AI expert Andrew Ng made a visit to Taiwan to take part in a series of forums, where he recently shared the stage with Audrey Tang, the Minister of Digital Affairs. They engaged in extensive discussions concerning AI models and regulations. Andrew Ng highlighted Taiwan's advantage in having a concentration of precision manufacturing industries, a strength that Silicon Valley lacks. In contrast, Audrey Tang advocated the growing prevalence of edge computing.

Back in 2018, Foxconn Group announced a collaboration with Andrew Ng's Landing AI to develop industrial AI applications. During his visit to Taiwan, Andrew Ng didn't delve into the progress of their collaboration but occasionally mentioned cooperative efforts between his team and other Taiwanese entities. These collaborations included the utilization of Landing AI software for aiding cosmic ray detection at the physics department of National Cheng Kung University, as well as partnerships with Taiwanese companies specializing in material classification. His other venture, DeepLearning.AI, also maintains engineering teams in Taiwan.

When asked about how Taiwan, with its semiconductor industry advantage, can advance AI development, Andrew Ng pointed out that many profound AI technological advancements have traditionally been centered in Silicon Valley. However, given the rapid pace of technological progress, existing norms are being rewritten, and now anyone can participate without being too late to the game. He believes that Taiwan should capitalize on its existing strengths, particularly its concentration of precision manufacturing industries, which Silicon Valley lacks. He also recommended that Taiwan leverage its robust hardware industry to progressively fortify its software sector.

Audrey Tang emphasized the escalating importance of data privacy and cybersecurity. With the advancement of Generative AI technology, previously effective techniques like de-identification and partial masking, used for safeguarding personal data, have become less potent. The government is tightening regulations related to personal data protection, and the Ministry of Digital Affairs is actively seeking extensive input regarding the ethical use of non-personal data. Cryptographic technologies like zero-knowledge proof are emerging as new trends, and they rely on advanced chip manufacturing techniques that facilitate the accelerated implementation of these encryption algorithms.

Audrey Tang also brought up the "infodemic" phenomenon during the COVID-19 pandemic, characterized by the rapid spread of false information, leading to societal panic. Social media platforms also shoulder a degree of responsibility in this context, and she stressed the necessity of transparency in algorithms and mechanisms for holding them accountable.

Regarding regulatory matters, Andrew Ng pointed out that in the United States, the government has limited control over private enterprises, and there are few avenues for both the government and the public to gain insights into technology-related risks. For instance, the Cambridge Analytica data manipulation incident only came to light years after it occurred. He also noted that in Silicon Valley, a small group of companies is currently utilizing Generative AI for potentially concerning projects. Therefore, he stressed the importance of carefully planning regulatory approaches for AI. Nevertheless, he cautioned that in certain countries, regulatory mechanisms that exclusively serve the interests of a select few can be more detrimental than a lack of regulation.

Both Andrew Ng and Audrey Tang discussed the utilization of open-source models from companies like Meta and subsequently fine-tuning them. Audrey Tang further underscored the imperative of democratizing AI and making it accessible for edge applications. She contended that rather than relying on a few large data centers with all the world's AI computing power, we should bolster technologies associated with edge AI. This approach would enable everyone to use AI as an auxiliary technology to address their specific needs. She emphasized that nobody would want to transmit sensitive data to the cloud, particularly without robust encryption safeguards. Consequently, edge computing and model fine-tuning are poised to become more prevalent. In this context, the capability to fine-tune models assumes paramount importance. Even if models exhibit biases and lack transparency, individual practitioners can find ways to enhance them. Audrey Tang mentioned that she personally employs open-source models for fine-tuning on her MacBook.

Audrey Tang's viewpoints align with recent industry conversations surrounding AI-powered PCs. Microsoft has introduced an array of AI assistant features in the Windows operating system, while Apple has not overtly emphasized AI in PCs. Nevertheless, former Google Taiwan CEO Lee-Feng Chien has posited that Microsoft and Apple should be at the forefront of the AI-powered PC wave. As software innovations accelerate, significant changes in PC hardware design are expected.