,

Taiwan AI Labs Founder Ethan Tu at NATO Summit

Photo Caption: Ethan Tu, founder of Taiwan AI Labs, was invited to Washington, D.C. to attend the NATO summit. He delivered a speech on the topic of information warfare during the 2024 European Parliament elections and engaged in discussions with the Special Competitive Studies Project (SCSP), an AI and national security think tank based in Washington.

 

[Taipei, Taiwan, July 12, 2024]  The manipulation of information by certain forces to influence election outcomes has become a global concern. On July 9, Ethan Tu, founder of Taiwan AI Labs, attended the NATO summit in Washington, D.C., where he delivered a speech addressing the issue of information manipulation during the 2024 European Parliament elections. Tu called for the establishment of trustworthy AI verification mechanisms to address the current crisis of information warfare facing democracies.

Taiwan AI Labs’ cognitive security tool, Infodemic, was utilized to analyze 26,011 social media battlefields and 335,045 major news stories across multiple platforms, including Facebook, YouTube, and X, from July 1, 2023, to June 24, 2024. According to Infodemic’s analysis, during the EU Parliament Election period, there were 20,041 troll group accounts displaying abnormal behavior, not operated by genuine users. These troll groups accounted for 12.58% of the discussions related to the EU Parliament Election.

The most influential troll groups were divided into two main categories. One group focused on Eastern European geopolitics and international military issues, such as the Russia-Ukraine war and NATO actions. The other group concentrated on attacking Western democratic systems, particularly the UK and the US, spreading negative narratives about national leaders and government decisions, aiming to undermine public confidence in democratic institutions. Further analysis revealed that these groups operated transnationally, also engaging in issues related to the US presidential election.

Troll Groups Align with Authoritarian State Media: European Parliament Faces Information Warfare Similar to Taiwan’s Presidential Election

The analysis indicated that these troll groups shaped public opinion by criticizing mainstream media for lack of impartiality, amplifying Russian propaganda, and attacking EU digital regulations aimed at combating misinformation and hate speech. This rhetoric potentially fosters far-right conservative ideologies and aims to instill doubt about the threats posed by the rise of extremism.

Moreover, during the same period, Infodemic used FedGPT to analyze 1,624 major news events, finding that 16.07% were related to Russian and Chinese state media. Notably, during the European Parliament election, on key issues such as the Russia-Ukraine war, energy security, digital regulation, immigration, and climate, the narratives of these troll groups aligned with those of Russian and Chinese state media, challenging the EU’s leadership narrative.

For instance, comparing July-August 2023 to September-October 2023, online attacks on the EU’s support for Ukraine increased from 30.5% to 34.97%. Negative public opinion on economic issues, particularly energy security, accounted for 22.04% of troll group narratives in September-October 2023, making it one of the largest battlegrounds during the election period.

Tu specifically pointed out that these opinion manipulation strategies are similar to phenomena observed by Infodemic during Taiwan’s 2024 presidential election. Troll groups exploited issues like military spending and energy security to erode public trust in the government and Taiwan’s international allies. Compared to traditional election propaganda, the information warfare strategies used by these troll groups are more sophisticated. They do not merely spread misinformation on single issues but weave targeted narratives across multiple countries and topics into everyday online interactions, deepening their impact on public opinion. This presents a significant challenge to democratic governance.

Global Stability Threats Extend Beyond Physical Battlefields: Developing Trustworthy AI Mechanisms to Counter Information Warfare

In the NATO summit speech, Taiwan AI Labs built on its recent engagement at the inaugural AI Expo for National Competitiveness held in May 2024 in the U.S. This event facilitated another critical exchange with the influential Washington think tank, the Special Competitive Studies Project (SCSP), focusing on the pressing issues of information warfare and national security.

Photo Caption: At the first AI Expo for National Competitiveness, Taiwan AI Labs was invited to participate. Founder Ethan Tu delivered a speech, sharing Taiwan’s experience in developing large open-source language models TAIDE and TAME, as well as further establishing the alliance-level FedGPT. He discussed applying AI technology to identify information manipulation and protect digital democracy.

 

Before the NATO summit, SCSP Executive Director Ylli Bajraktari published an article on Project Syndicate, a leading global commentary website, titled “AI-Augmented Disinformation Is NATO’s New Battlefield.” Bajraktari emphasized that the threats to global stability have surpassed traditional military domains. NATO member states must directly respond to cognitive warfare initiated by hostile authoritarian regimes. Specific measures include investing in tools capable of identifying content authenticity, such as large language models (LLMs) and AI classifiers, to detect AI-generated or altered content and identify malicious activities by adversarial factions on digital platforms, thereby reducing their influence on the public. Additionally, given that information manipulation often transcends national borders, NATO should ally with governments, civil organizations, and private companies to establish early warning systems and collaborate to counter large-scale information warfare.

Ehtan Tu stated that 2024 is a global election year, and echoing Ylli Bajraktari’s views, it is crucial to collaborate with reliable partners to establish AI mechanisms that expose information manipulation tactics by coordinated online accounts, thereby enhancing the resilience of democratic systems against threats such as disinformation and ideological polarization. Taiwan AI Labs’ Infodemic platform is an AI-driven cognitive security platform designed to detect modern information warfare practices, track the spread of false information, and reveal the targets, tactics, and strategies employed by adversaries that may impact national security.

/by
,

Global Debut of Traditional Chinese Financial GPT: Taishin Brain Launches

[Taipei, Taiwan, July 5, 2024]Taiwan AI technology is poised to revolutionize the global financial industry. Today, Taishin Bank and Taiwan AI Labs announced the official launch of “Taishin Brain,” a groundbreaking financial GPT system tailored for the financial sector with Traditional Chinese as its core language. This project marks a new chapter in Taiwan’s financial AI innovation.

“Taishin Brain” integrates the advanced FedGPT, a Traditional Chinese large language model, within a framework of trusted and responsible AI. Unlike conventional GPT models, it is meticulously trained using Taishin Bank’s internal data and knowledge, including financial regulations and expertise, ensuring it meets stringent global regulatory standards. This makes “Taishin Brain” the world’s first Traditional Chinese language model to address the highly regulated financial sector’s demands.

With specialized training in financial expertise, “Taishin Brain” delivers precise, intelligent responses even to ambiguous queries, setting a new standard for humanized AI in the financial domain. Leveraging federated learning (FedGPT), it ensures data privacy and security while utilizing Taishin Bank’s professional data to meet the unique needs of financial institutions. This integration enhances the professionalism and efficiency of human customer service.

Through advanced AutoML capabilities, “Taishin Brain” autonomously learns and continuously improves. It connects with Taishin Bank’s internal financial systems via AutoCopilot to streamline processes effectively.

“Taishin Brain” is a pioneering leap in financial technology, positioning Taiwan’s financial industry at the forefront of global intelligence and significantly boosting the competitiveness of Taiwan’s financial market.

 

FedGPT: Taiwan’s First Enterprise-Dedicated Traditional Chinese GPT

While ChatGPT interacts, analyzes, and provides information across various data formats, its training in the Chinese domain has primarily focused on Simplified Chinese, with limited financial expertise. In contrast, FedGPT, developed by Taiwan AI Labs in 2023, is designed for Traditional Chinese and specialized fields like finance, healthcare, and media.

FedGPT’s federated learning model allows institutions to keep data in-house while training on diverse document types, ensuring compliance with the Financial Supervisory Commission’s AI guidelines. Users can select knowledge bases tailored to their specific institution and domain, resulting in more accurate financial responses.

 

Technological Innovation Meets Financial Expertise

“Taishin Brain” leverages internal technical fine-tuning and automated machine learning, enabling effective training with minimal data and resource usage. The system continuously updates and optimizes based on data, ensuring accurate and reliable responses in the financial sector.

 

Building a Comprehensive Knowledge Base

The first phase of “Taishin Brain” focuses on establishing an extensive knowledge base, including financial regulations, legal knowledge, and Taishin Bank’s corporate data. This resource will significantly enhance internal information retrieval and employee efficiency.

 

Commitment to Information Security and Privacy

“Taishin Brain” uses AI algorithms to prevent biases, hallucinations, and harmful content. Secure classifiers filter inappropriate material, and data governance ensures legality, compliance, and explainability, enhancing model training efficiency and data privacy.

 

Practical Applications: Smart Customer Service and Knowledge Management

“Taishin Brain” excels in smart customer service and knowledge management. Unlike traditional chatbots, its generative AI integrates information and provides intelligent, accurate responses, addressing customer service pain points, reducing labor costs, and improving corporate efficiency.

Ethan Tu, founder of Taiwan AI Labs, stated, “Our goal is to develop a trusted and responsible AI that is human-centered and protects privacy. As the world considers integrating ChatGPT into financial institutions, we are addressing labor and time cost challenges, providing intelligent customer service solutions both internally and externally.”

Taishin Bank views “Taishin Brain” as a significant breakthrough in Taiwan’s financial technology sector, enhancing service quality and customer experience. In collaboration with Taiwan AI Labs, Taishin Bank will continue to develop this AI brain for its employees and customers, offering intelligent, convenient solutions across more domains in the future.

 

About FedGPT (https://yating.tw/zh-hant/fedgpt-landing)

FedGPT is an on-premises large language model technology developed by Taiwan AI Labs. It is built using Taiwan’s own LLM (Large Language Model) technology, trained with over 60 billion tokens of Traditional Chinese corpus. The model emphasizes providing grounded responses based on internally established knowledge bases or verified through federated learning alliances, ensuring a trustworthy AI model. The technology adheres to principles of trusted and responsible AI, with a strong emphasis on information security.

 

About Taiwan AI Labs

Taiwan AI Labs (https://ailabs.tw)  , operating under the guiding principle of trusted AI, effectively integrates generative AI into a range of solutions, including smart healthcare, federated learning, AI virtual hosts, future scenario generation studios, and news analysis platforms. In the field of human-computer interaction, they develop technologies for AI-generated music and art, speech recognition, and semantic understanding. A major application addresses the global concern of disinformation and fake news, analyzing the context and patterns of information manipulation. These achievements have sparked lively discussions internationally and have been frequently recognized and published in authoritative international journals.

 

Taiwan AI Labs continues to lead cutting-edge AI research, aspiring to form a national AI team by integrating talent, resources, and the industrial chain. Their goal is to establish a Taiwan Federated Learning Industry Alliance and promote federated solutions from both public and private clouds to a global audience.

/by