,

AI Experts Embedded in the Enterprise: FedGPT AgentTeam Brings Agentic AI into Real-World Workflows

Photo Caption: From left to right in the front row — JiaHsin Huang, General Manager of Industry Solutions at Taiwan AI Labs; Jon Wang, Managing Director and President for HPE Taiwan; Ethan Tu, Founder of Taiwan AI Labs; Brandon Lin, Director of Strategic Marketing, Imaging, Care Pathways & AI Solutions at GE HealthCare; and Benson Tu, Director of Core Platform Products at Taiwan AI Labs.

While generative AI is now widespread, enterprises still face major challenges: high deployment costs, stringent data privacy concerns, and fragmented internal workflows that hinder the effective use of AI in daily operations. FedGPT AgentTeam addresses these pain points with an on-premise, trustworthy multi-expert small-model architecture. Data stays on local infrastructure, enabling secure, cross-departmental collaboration and multimodal AI workflows that overcome the limitations of traditional tools.

“FedGPT AgentTeam is tailored to the nuances of Traditional Chinese used in Taiwan and designed with a multi-expert, multi-modal, and multi-center framework,” said Ethan Tu, Founder of Taiwan AI Labs. “It allows enterprises to rapidly deploy specialized AI Agents that integrate seamlessly with both internal systems and external tools. From customer service to HR, marketing, manufacturing, and IT, these agents learn internal knowledge and automate complex tasks—turning AI into a true driver of productivity and growth.”

The Rise of Agentic AI: FedGPT AgentTeam Powers End-to-End Workflow Automation

According to The Economist, 2025 marks a turning point for Agentic AI. Gartner predicts that by 2029, 80% of customer service operations will be autonomously handled by AI Agents. As this shift accelerates, FedGPT AgentTeam is emerging as Taiwan’s leading Agentic AI deployment platform, enabling enterprises to scale from single-point AI tools to holistic, organization-wide systems.

The version name “AgentTeam” reflects the platform’s core value: quickly assembling a team of specialized AI Agents tailored to each enterprise’s needs and knowledge base, working together to complete complex business processes.

Three Core Capabilities to Enhance Business Intelligence and Efficiency

FedGPT AgentTeam combines a domain-specific large language model (LLM) with advanced multimodal AI modules including ASR (automatic speech recognition), TTS (text-to-speech), and VLM (vision-language models). This enables it to process voice, images, and text seamlessly. Key features include:

  1. AgentTeam RAG

    Supports diverse formats and sources,including images, videos, charts, audio, and text,for intelligent knowledge retrieval and generation. It intelligently assesses the complexity of each query and selects the most appropriate problem-solving strategy.

    Example: When given a company’s stock chart and related news articles,FedGPT AgentTeam can analyze correlations between public statements and stock movements to generate actionable insights.
  2. AgentTeam Flow

    A built-in process orchestration tool with a low-code/no-code interface. Employees can trigger multi-agent workflows using natural language—no engineering background required.

    Example: In Taiwan AI Labs’ internal project approval process, tasks like requirement analysis, feasibility evaluation, and cost assessment are now fully automated using Flow.
  3. AgentTeam Tuning

    Supports enterprise-specific fine-tuning of AI models across multiple modalities, adapting to industry-specific terminology and communication styles.

    Example: In healthcare, the platform learned medical staff’s use of informal English abbreviations, enabling accurate transcription and workflow automation across departments.

FedGPT AgentTeam is designed for on-premises deployment, ensuring sensitive data never leaves an organization’s secure environment, which is ideal for highly regulated sectors such as finance, healthcare, and manufacturing.

FedGPT AgentTeam is powered by domain-specific language models shaped by Taiwan AI Labs to align with local linguistic and enterprise needs.Built for adaptability, the model can be rapidly fine-tuned to reflect the specific linguistic, cultural, and operational contexts of each enterprise. In a benchmark assessing localized understanding, the model achieved a Taiwan cognition score of 81.4 out of 100, substantially higher than Chinese models such as Qwen (44.3) and DeepSeek (38.7),demonstrating its strong capability in interpreting and generating content aligned with Taiwan’s unique language environment.

On the “Chinese ideological bias” metric—where lower scores indicate reduced susceptibility to Chinese state-aligned narratives—FedGPT’s score was significantly lower than both Chinese and Western open-source models. This positions the platform as a safer and more contextually appropriate AI solution for deployment in Taiwan’s public sector and sensitive industry environments, minimizing the risk of propagandistic phrasing or narrative misalignment.

To build this domain-specific capability, FedGPT AgentTeam models underwent domain-adaptive pretraining on over 100 billion tokens of Taiwan-specific professional data, requiring just 4,600 NVIDIA H100 GPU hours—an exceptionally efficient process for achieving contextual alignment. The platform is also optimized for enterprise performance: the standard version runs both AgentTeam Flow and AgentTeam RAG on a single Nvidia H200 GPU, dramatically reducing hardware requirements and energy consumption compared to traditional LLMs.

With sustainable infrastructure and predictable total cost of ownership (TCO), the platform offers a cost-effective alternative to token-based pricing models of cloud AI services.

Seamless System Integration with MCP and A2A Protocols

FedGPT AgentTeam supports two open standards that power agent collaboration and integration:

MCP (Model Context Protocol): A secure interface that facilitates intermediary-based access to external applications such as Gmail, Slack, and Google Drive. It converts diverse data formats into AI-readable inputs—think of it as a USB adapter for AI, enabling plug-and-play interoperability.

Photo Caption:MCP can be thought of as a USB adapter for AI, enabling plug-and-play interoperability.

A2A (Agent-to-Agent): A communication protocol for AI Agents to share information, sync progress, and coordinate decision-making. It mirrors how human experts consult each other.Agents can request input from more specialized peers, enabling collaborative problem-solving for complex tasks.

For quality control, the system includes a Human-in-the-Loop mechanism, allowing users to set review checkpoints. It also integrates ReAct (Reasoning and Acting) logic, enabling agents to self-evaluate their outputs at each step, reducing the risk of error accumulation in multi-stage workflows.

Broad Adoption Across Key Industries

FedGPT AgentTeam is already being used in healthcare, finance, education, and government. Early adopters include Hualien Tzu Chi Hospital, Fu Jen Catholic University Hospital, Dong Hwa University, and Taishin Bank.

Looking ahead, the platform is poised to transform high-volume, repetitive business functions in HR, admin, customer service, legal, marketing, and IT. Applications range from policy search and document review to leave approvals, marketing content generation, legal Q&A, and code generation.

Unlike traditional ERP systems or RPA tools, FedGPT AgentTeam enables real-time, natural-language-triggered workflows that adapt to multimodal data. With its intuitive, language-based interface, the platform has the potential to redefine enterprise architecture, eventually replacing parts of legacy ERP systems and bringing AI into the core of daily operations.

/by
,

Tzu Chi University and Paramitas Education Association Join Forces to Deploy FedGPT AgentTeam, Developed by Taiwan AI Labs, Enhancing AI-Powered Learning Support

[June 18, 2025, Taipei] — In response to the rapid advancement of artificial intelligence and its transformative impact on education, Tzu Chi University and the Paramitas Education Association officially signed a memorandum of understanding (MOU) today to integrate FedGPT AgentTeam, an intelligent learning platform developed by Taiwan AI Labs. This collaboration marks a significant step toward bridging educational innovation with humanistic values by introducing AI-powered tools into teaching, research, and campus operations.

The MOU was signed by Dr. Ingrid Y. Liu, President of Tzu Chi University, and Mr. Wen-Chin Huang, Chairman of the Paramitas Education Association, with Mr. Hong-Wen Chang, Chairman of Wizigo Co., Ltd., Ethan Tu, Founder of Taiwan AI Labs, and Chaohwei Huang, Chief Content Officer, witnessing the historic partnership. The collaboration focuses on four key areas: platform deployment and technical support; curriculum integration and pedagogical innovation; joint project development and results sharing; and ethical standards in data usage.

The Paramitas Education Association has long championed personalized education solutions using AI to support self-directed learning. By leveraging data from students’ learning histories, the organization has developed student-centered teaching frameworks grounded in evidence-based practices.

FedGPT AgentTeam is an agentic AI platform developed by Taiwan AI Labs, designed to support on-premise deployment, multimodal data processing, and highly automated workflows. While widely applicable across enterprise scenarios, the platform can be effectively adapted for educational use—offering capabilities such as intelligent knowledge retrieval, streamlined administrative processes, and enhanced teaching support.

The system connects with internal and external resources via the Model Collaboration Protocol (MCP) and Application-to-Application (A2A) integration, establishing a secure and trustworthy AI ecosystem. Core capabilities include:

  • AgentTeam RAG: Facilitates cross-source, multimodal knowledge retrieval by integrating internal documents, curricula, regulations, and educational media to provide real-time, reliable academic support.
  • AgentTeam Flow: Enables low-code process design, helping schools automate administrative and academic workflows—from academic records and cross-department collaboration to approvals—improving efficiency and reducing manual workload.
  • AgentTeam Tuning: Allows customization of the AI model to understand domain-specific language and context, enhancing AI performance across teaching and administrative scenarios.
  • Human-in-the-Loop Mechanism: Embeds a human review process throughout to ensure trustworthiness, accuracy, and compliance with data ethics in educational settings.

With support from both the Paramitas Education Association and Taiwan AI Labs, Tzu Chi University will implement FedGPT AgentTeam to enhance its Chinese-language capabilities in natural language understanding and academic assistance—empowering faculty and students to conduct more efficient and in-depth research and learning.

A key feature of the initiative is the rollout of a “24/7 AI Teaching Assistant”, designed to provide continuous learning support. The assistant offers immediate responses to student inquiries and delivers personalized guidance and extended resources based on individual learning levels—particularly beneficial for students needing additional support to bridge gaps and boost motivation.

“AI is not here to replace people, but to support them,” said Dr. Ingrid Y. Liu, President of Tzu Chi University. “Our AI assistant was designed to offer every student timely companionship and guidance, making learning accessible anytime, anywhere.”

Another major focus of the partnership is the “Tzu Chi and Social Development Research Project”, which includes the creation of three core knowledge bases: the Tzu Chi History and Culture Database, the Humanitarian Media Archives, and the Tzu Chi Global Events Knowledge Base. These resources will be powered by the FedGPT AgentTeam platform and aim to help faculty and students explore Tzu Chi’s contributions to humanitarian relief, cultural communication, and global engagement—fostering interdisciplinary and intergenerational research.

By leveraging AI, students will gain a deeper understanding of Tzu Chi’s mission while improving their abilities in research writing, data organization, and critical thinking—developing global perspectives alongside a profound sense of cultural empathy.

“AI must be embraced in the future of education, but never at the expense of its core values,” emphasized Mr. Wen-Chin Huang, Chairman of the Paramitas Education Association. “This partnership with Tzu Chi University is not just about sharing technology—it’s about co-creating values.”

Ethan Tu, Founder of Taiwan AI Labs, added: “The true value of an AI teaching assistant lies in its ability to understand the needs of the school and evolve with the learning environment. FedGPT AgentTeam is like a tireless digital assistant—always learning, always ready. It helps students with questions, supports teachers and administrators, and allows human resources to focus on meaningful, human-centered education.”

Looking ahead, Tzu Chi University will continue to integrate AI into its curriculum—guiding students from understanding AI, to applying it, and ultimately creating with it. Through practical applications such as the AI Teaching Assistant and research-oriented knowledge bases, the university remains committed to its founding values of compassion, joy, and selfless giving, while nurturing future professionals equipped with both empathy and forward-thinking innovation.

To further extend the impact of this initiative, the Paramitas Education Association will host a national education forum titled “Empowering Adaptive Learning with AI” on July 19, 2025, inviting educators across Taiwan to explore the future of collaborative, AI-enhanced learning.

[More Information]
 Tzu Chi University: https://www.tcu.edu.tw/
 Paramitas Education Association: https://twpea.org/
 FedGPT AgentTeam Platform: https://fedgpt.cc

 

/by
,

Taiwan AI Labs Recognized as a Trusted AI Partner Aligned with EU Regulations; Launches First National-Level Federated Learning Collaboration Between Taiwan and Europe

[Taipei, May 15, 2025] Taiwan AI Labs was recently recognized as a Gold Partner at Machine Learning Prague 2025, one of Europe’s most influential applied AI conferences, held from April 28 to 30 at the O2 Universum in Prague, Czech Republic. The Lab was formally recommended by the Czech National AI Hub as a trusted international partner, with its technological direction aligned with the EU’s Artificial Intelligence Act (AI Act) and General Data Protection Regulation (GDPR).

The event gathered thousands of data scientists, engineers, and industry leaders worldwide to explore practical applications of AI and discuss key challenges. This year’s agenda spotlighted AI in healthcare and finance, information security, and the advancement of trustworthy AI technologies.

Machine Learning Prague introduced Taiwan AI Labs as Asia’s first open AI research institute, noting its deep alignment with the EU’s core AI values. The Lab’s work reflects a shared vision in four key areas:

  1. Federated Learning & Data Privacy
     Enabling AI training across decentralized data without exposing sensitive personal information.
  2. Healthcare AI & Ethical Governance
     Developing clinically validated, explainable systems under ethical oversight.
  3. Civic Tech & Misinformation Resilience
     Creating tools to counter disinformation and strengthen democratic institutions.
  4. Transparency & Open Innovation
     Publishing research outcomes, promoting open-source models, and embedding accountability mechanisms.

In response to the global momentum around digital rights and data protection, Taiwan AI Labs developed FedGPT—a platform that integrates federated learning with generative AI. Designed to meet stringent privacy and compliance standards, FedGPT allows institutions to train AI models internally without transferring sensitive data to external clouds or third parties. Its architecture supports cross-institutional and cross-border collaboration, making it ideal for sectors such as healthcare, finance, and public administration.

Ethan Tu, Founder of Taiwan AI Labs, remarked:“Trustworthy technology is Taiwan’s strength. In today’s landscape, regulatory compliance is not a burden—it’s a prerequisite for participating in the global AI ecosystem.”

Following its participation in the AI Action Summit in France last year, Taiwan AI Labs has now been invited by the Czech Ministry of Health and national medical institutions to co-launch a collaborative initiative on federated learning and generative AI. This joint effort will support the Czech healthcare system in adopting AI solutions that prioritize privacy and ethics—marking a significant step in Taiwan’s AI technology being applied in real-world European contexts.

Tu added,“Every industry, everywhere, will need trusted AI solutions. Taiwan’s native approach—rooted in responsibility, compliance, and control—aims to be a reliable choice for our global partners.”

About Taiwan AI Labs

Founded in 2017, Taiwan AI Labs is a mission-driven, privately funded open AI research institute based in Taipei and Kaohsiung. The Lab develops AI technologies grounded in trust, privacy, and human-centric design, delivering solutions across smart healthcare, federated learning, AI-generated media, and disinformation detection. Its innovations in human-AI interaction—including AI music generation, speech recognition, and semantic understanding—have driven global conversations and earned recognition from top international journals.

The Lab continues to lead on the global frontier, rallying Taiwan’s talent, resources, and industries to form a National Federated Learning Alliance—bringing federated AI deployment from both public and private clouds to the world stage.

Media Contact
 Ellie Chen
 Public Relations, Taiwan AI Labs
 wanhsin.chen@ailabs.tw

/by
,

Taiwan AI Labs and Tzu Chi Culture and Communication Foundation Partner to Launch Custom Media GPT: Harnessing Technology to Spread Compassionate Values

[Taipei, April 21, 2025] — Taiwan AI Labs and the Tzu Chi Culture and Communication Foundation today announced the signing of a memorandum of understanding to co-develop the nation’s first GPT-powered media solution, built on a domain-specific knowledge base and tailored for video and audio applications.

The service will aggregate over 100,000 videos and media assets collected by Da Ai Television since its founding in 1998. Leveraging AI technologies, it will automatically generate video summaries, scene logs, and continuously learn from data to assist production teams in creating scripts, manuscripts, and other innovative content formats—marking a major milestone in the digital transformation of Taiwan’s media asset management.

This collaboration is a pivotal step in realizing Tzu Chi’s vision of “Spreading the Dharma Through Technology.” For more than two decades, the Foundation has shared messages of truth, kindness, and beauty to over 136 countries and regions through satellite broadcasting, radio, websites, print publications, and social media. In the age of AI, this partnership focuses on revitalizing nearly 60 years of Tzu Chi’s archival content, extracting new thematic insights through AI while allowing media professionals to focus on producing emotionally resonant content—an embodiment of collaborative goodness between humanity and technology.

Duan-Jen Wang, CEO of the Tzu Chi Culture and Communication Foundation, emphasized that the mission is not merely to adopt generative AI, but to develop a large language model that deeply understands and conveys the spirit of Tzu Chi. “It may not yet feel weighty today,” he said, “but in the long run, this initiative bears witness to our times, documenting human history and preserving Tzu Chi’s legacy.”

Eric Yao, Director of Strategic Alignment and Development at the Foundation, noted that this collaboration is more significant than imagined. “We’re entering an era of Four Intelligences,” he said: Awakening Intelligence, without which there is no purification of the mind; Aesthetic Intelligence, which transforms beauty into impactful energy; Artificial Intelligence, which becomes more beneficial to humanity when guided by awakening and aesthetic sensibilities; and Altruistic Intelligence, the ultimate goal that Master Cheng Yen has long sought to realize. This partnership, he added, aims to integrate all four.

Ethan Tu, Founder of Taiwan AI Labs, shared that the teams are developing an AI agent with multimodal and multi-expert capabilities. It will recognize the who, what, when, where, and how of audiovisual content and support Da Ai’s scene-logging workflows, helping media professionals rapidly retrieve relevant information from vast archives.

In a field that prioritizes truth, AI must be professional and trustworthy. Tu pointed out that mainstream open-source models are akin to university interns with general knowledge but little domain-specific expertise. Training such models with sensitive internal data on external cloud services poses risks. In contrast, Taiwan AI Labs’ FedGPT was built under principles of trustworthy and responsible AI governance. Like a freshly minted PhD joining a company full-time, FedGPT can be retrained internally without sharing data externally, learning efficiently and securely through federated learning—loyal to the organization, without leaking sensitive information.

In today’s competitive AI landscape, FedGPT stands out as an internal expert model aligned with corporate sustainability and ESG values, consuming minimal energy while offering high efficiency and deep specialization. At Tzu Chi, FedGPT will act as a diligent new team member, boosting productivity in content production.

FedGPT supports Mandarin, Taiwanese, and English speech recognition, handling over 1.7 million hours of processing per year. It understands local linguistic nuances such as the Taiwanese idiom “Tiām-tiām tsia̍h sann uánn-kong puànn” (Still waters run deep). Unlike international models dependent on the cloud, FedGPT is equipped with Taiwan-specific anti-bias datasets and a Guardian mechanism to curb hallucinations and biases, offering more culturally attuned and accurate results.

Three Core Collaboration Areas for Smart Media Transformation

The collaboration focuses on three core AI services:

  1. Scene-Logging GPT for Tzu Chi: Built on Taiwan AI Labs’ multimodal, multi-expert FedGPT platform, this tool will automatically interpret video content and build a structured media knowledge base with details like people, events, locations, and objects. 
  2. Topic-Guided Assistant GPT: After selecting a theme, production teams can use FedGPT to curate related historical footage, streamlining the editorial workflow and enhancing real-time content delivery. 
  3. AutoML Platform for Active Learning: Beyond pre-training on domain-specific data, FedGPT will leverage user feedback and internal video content to continuously refine Tzu Chi’s custom language model, improving analysis and productivity.

Ethan Tu emphasized that the year-long collaboration has helped Taiwan AI Labs deeply understand the real-world challenges of media workflows. FedGPT integrates a large language model (LLM), vision-language model (VLM), automatic speech recognition (ASR), enterprise knowledge base (AutoKB), and task assistant (AutoCopilot). Its training spans trillions of Traditional Chinese and English tokens, over 5,000 hours of Taiwanese speech data, and tens of thousands of annotated faces and clothing images—enhancing its ability to recognize local visual and audio patterns.

Even within the same person, changes in angle, hairstyle, or attire can challenge recognition systems. FedGPT tackles this with precise object recognition. Users can define targets—Master Cheng Yen’s robe, Tzu Chi’s signature blue-and-white volunteer uniforms, or specific volunteers’ faces—and retrieve all relevant images or footage instantly. It achieved 94.81% top-1 facial recognition accuracy in Google Images benchmark tests, demonstrating robust performance in complex environments.

For instance, if a team urgently needs footage of Master Cheng Yen’s travels, FedGPT can swiftly locate all relevant clips—even when masks and identical clothing obscure identities—saving countless hours of manual review.

FedGPT offers three levels of scene-log granularity:

  1. Shot-by-shot visual topic identification. 
  2. Segment-based thematic analysis (e.g., opening, interview, product demo). 
  3. Full-video summaries with thematic essence. 

This dramatically reduces the manual labor of drafting scene logs or ensuring content quality. For teams handling dozens of clips daily, simply uploading media to FedGPT yields structured metadata like: “The reporter is conducting a street interview,” or “Cut to Reporter A’s live segment.”

The team also enhanced FedGPT’s contextual understanding of visuals, speech, and subtitles. For example, in disaster footage, it can be discerned that collapsed structures and overturned vehicles indicate a post-earthquake recovery scene, not a single accident.

Tu concluded, “AI grows through learning, and we’ve learned so much from Tzu Chi. Now it’s time to teach GPT to learn proactively—to speak a language aligned with Tzu Chi’s values. In this irreversible transformation of the media industry, this step we’re taking together is not just Taiwan’s first—it’s a world first. Tzu Chi is our first and best teacher.”

Today, many production tasks still rely on manual labor—especially scene logging. With FedGPT, challenges like parsing footage, logging scenes, and labeling dialogue or visuals can now be automated. It can adapt output formats to suit various teams and projects, semantically segmenting and summarizing content to ease editorial burdens.

With its ability to comprehend and analyze massive archives, FedGPT enhances search and editing workflows. It also learns from user feedback, gradually replacing traditional manual training processes. In the future, Tzu Chi’s production teams will interact directly with their custom GPT—not only improving efficiency but also generating content in Tzu Chi’s unique tone, further spreading the organization’s spirit of “united in heart, regardless of religion.”

/by

Taiwan AI Labs and Phison Electronics Join Forces to Deliver FedGPT All-in-One Solutions with Multimodal and Multi-Expert Capabilities

[March 27, 2025] At AI EXPO Taiwan 2025, Ethan Tu, Founder of Taiwan AI Labs, and Pua, Khein-Seng, CEO of Phison Electronics, engaged in an in-depth dialogue on the future of artificial intelligence. The two leaders announced a strategic collaboration to jointly promote an integrated hardware-software solution—the FedGPT All-in-One System—featuring multimodal and multi-expert capabilities. This solution aims to empower enterprises with a fully on-premise, energy-efficient, and trustworthy ReGenerative AI system built on Taiwan AI Labs’ aiDAPTIV+ technology platform. Designed with cost-effectiveness, ESG compliance, Taiwan’s domain expertise, and global AI regulations in mind, the solution brings practical generative AI deployment within reach for businesses.


photo caption: Ethan Tu (right), Founder of Taiwan AI Labs, and Pua, Khein-Seng (left), CEO of Phison Electronics, in discussion at AI EXPO Taiwan 2025.

Ethan Tu emphasized that with advances in AI infrastructure in 2025, enterprises can now leverage foundational models trained by Taiwan AI Labs to build custom, domain-specific expert models using the enterprise-grade federated GPT platform, FedGPT.

He stated, “2025 marks the first year of ReGenerative AI. With Taiwan’s strengths in both software and hardware, FedGPT enables organizations to train multimodal, multi-expert models using internal data—no internet access or external subscriptions required. It meets regulatory, ESG, and cost requirements, empowering companies to build high-ROI, domain-specialized AI brains. Through our collaboration with Phison, we’re advancing technologies that allow for full-parameter model training and simultaneous service delivery—on a single device.”

Following the open-source movement and regulatory governance in 2024, sustainable AI is set to become the norm in 2025. Developed by Taiwan AI Labs, FedGPT is a federated generative AI platform tailored for Traditional Chinese. Trained on 100 billion Traditional Chinese data points, plus 15 billion in medical datasets and 1 billion in business, finance, and regulatory materials, FedGPT is designed to be lightweight and energy-efficient, while delivering top-tier performance across multiple professional domains. It provides a practical alternative to resource-intensive models, offering accurate, domain-aware outputs tailored to enterprise needs.

Pua, Khein-Seng noted that AI hardware will increasingly move toward high performance, low power consumption, and scalability. Phison’s patented aiDAPTIV+ architecture seamlessly integrates hardware and software to lower the barrier for on-premise AI deployment while ensuring data privacy. By integrating SSDs into the AI compute framework and using proprietary technology to expand DRAM capabilities, organizations can use standard GPUs to fine-tune large language models—dramatically reducing deployment costs and ensuring private, edge-based AI training.

He also highlighted a gap in industry familiarity and AI talent, which FedGPT is designed to address. Ethan Tu described FedGPT as a standardized toolkit for industries to build their own domain-specific models. He likened FedGPT’s pretrained models to PhD graduates from the “Taiwan AI Labs Institute” who then specialize in finance, medicine, or other sectors. These models continue to evolve post-deployment by learning from internal data, transforming from “graduates” into internal experts, delivering high-value outcomes for organizations.

Since its launch in 2024, FedGPT has been successfully deployed across finance, healthcare, education, and the public sector. Notable use cases include Hualien Tzu Chi Hospital (AI-assisted hospital workflows), Taishin Financial Holdings (“Taishin Brain”), and National Dong Hwa University (conversational campus assistant NDHU-GPT).

Key Areas of Technical Collaboration Between Taiwan AI Labs and Phison Electronics:

  1. Promoting Sustainable, Scalable AI Across Industries
     Combining Taiwan’s large-scale foundational models with Phison’s aiDAPTIV+ hardware and Taiwan AI Labs’ AutoML and AutoKB tools, this collaboration enables enterprises to create expert models and internal knowledge bases entirely offline, facilitating scalable AI deployment across all sectors.
  2. Advancing Data Justice and Federated Learning
     The Federated Platform (FP) in FedGPT allows multiple departments or organizations to jointly train models without sharing raw data, preserving data autonomy and protecting sensitive information, including trade secrets and customer data, while enabling cross-domain collaboration.
  3. Meeting Global Regulatory and Security Standards
     Built by Taiwan-based teams, the joint solution complies with global standards including the EU AI Act, GDPR, and ISO security frameworks, offering trustworthy AI solutions for the future global supply chain.
  4. Optimizing Enterprise AI Speed and Cost
     The integration of FedGPT and aiDAPTIV+ supports cost-effective, all-in-one, on-premise deployment—no specialized AI talent or external services needed. This lowers adoption barriers, accelerates deployment timelines, and allows for future flexibility and scaling.

Taiwan AI Labs and Phison will continue developing industry-specific AI solutions, driving full-scale adoption of generative AI technologies across Taiwan’s industrial landscape.


Photo caption: Taiwan AI Labs and Phison Electronics announce their strategic partnership at AI EXPO Taiwan 2025, launching an all-in-one FedGPT solution for trusted, sustainable, and regulation-compliant enterprise AI applications.

 


Photo Caption: Ethan Tu (right), Founder of Taiwan AI Labs, and Pua, Khein-Seng (left), CEO of Phison Electronics, pictured at AI EXPO Taiwan 2025.

About Taiwan AI Labs

Taiwan AI Labs is committed to building trustworthy AI. The Lab integrates generative AI into smart healthcare, federated learning, AI-generated broadcasting, immersive scene simulation, and news analysis platforms. Its AI music and speech technologies address critical global issues such as misinformation and have been featured in international academic journals and media.

Taiwan AI Labs continues to lead global AI research, working to assemble Taiwan’s “AI national team” by uniting talent, resources, and industries through federated learning solutions across public and private clouds.

About Phison Electronics Corp.

Phison Electronics Corp. (TPEX: 8299) is a global leader in NAND flash controllers and storage solutions. Offering comprehensive solutions from IP licensing and chip design to system architecture and product integration, Phison’s offerings span SSD (PCIe/SATA/PATA), eMMC/UFS, SD, and USB interfaces. Its products serve consumer, enterprise, industrial, and automotive markets. Phison is an active board member of SDA, ONFI, and UFSA, and contributes to JEDEC, PCI-SIG, MIPI, NVMe, and IEEE-SA standards.

Media Contacts

Taiwan AI Labs PR
 Ellie Chen
 wanhsin.chen@ailabs.tw

Phison Electronics Corp. Spokesperson
 Kuo-Ting Lu
 kuoting_lu@phison.com

 

/by
,

Hualien AI Music Lab: Musicians and AI Co-Create Songs Showcasing Taiwan’s Unique AI Pop Music Aesthetic

[March 14, 2025] The Hualien AI Music Lab held its Opening Showcase today, featuring prominent musicians such as legendary folk singer and Lab’s Operations Director Li Jianfu (Li Jian Fu), Music Director Ado’ Kaliting Pacidal, as well as Suming Rupi and Lim Giong. The event unveiled four AI-assisted compositions inspired by “Whales, Ocean, Flowers, and Birds”—iconic elements of Hualien’s natural landscape. These works merge Taiwan’s rich cultural diversity with AI technology, symbolizing AI’s integration into everyday life beyond its cold, mechanical image.

Dubbed the “Godfather of Folk Music”, Li Jianfu was an early adopter of AI in music, having played a key role in developing AI i-Broadcast, the world’s first officially licensed Mandarin AI singing platform. At today’s event, Li guided the AI in recreating youthful renditions of Fan Yiwen and Li Mingde’s voices and visuals. Together, they performed “Drifting Bird” (composed by Jin Tiezhang and written by Li Shuping), blending real and AI-generated vocals to create a dreamlike fusion of past and present.

Renowned composer and film scorer Lim Giong, known for fusing tradition with technology, crafted an exclusive soundtrack for the “Creative Otherworld” exhibition space at the Lab. Infused with electronic elements, the music evokes the flow and vastness of the ocean, enhancing the immersive audiovisual experience. Visitors’ AI-generated lyrics are projected onto the walls, floating alongside cosmic whales gliding through a starry universe.

Renowned Amis singer-songwriter Ado’ Kaliting Pacidal, dedicated to preserving Indigenous culture, presented her original song “A Poem from the Ocean to the Island”. The song illustrates the interdependent relationship between the island and the sea—the Pacific, like a mother’s apron, embraces Taiwan as waves kiss Hualien’s shores, inspiring the songs of ancestors. The piece incorporates the Amis call-and-response singing style, with Ado’ leading and her AI-generated voice responding, symbolizing AI’s role in empowering artists through collaboration.

Suming Rupi, an Amis singer from Dulan, Taitung, drew inspiration from Hualien’s “Flowers” to compose “If I Were a Flower”, a heartfelt song performed alongside an AI-generated vocal track. The accompanying AI-generated music video incorporates Hualien’s signature natural elements—flowers, stones, the ocean, mountains, pine trees, and rice fields—bringing the island’s emotional and cultural essence to life in a unique visual and auditory experience.

The Hualien AI Music Lab, powered by Taiwan AI Labs, integrates Taiwan’s native AI creation tools. At its core is AI Music FedGPT, a GPT model specifically trained for Traditional Chinese pop music, with an extensive dataset of 100 billion Taiwanese linguistic records. Historically, AI-generated vocals have predominantly used Mandarin models, but through collaborations with Indigenous artists Ado’ Kaliting Pacidal and Suming Rupi, the Lab has successfully developed authentic AI-generated Indigenous vocals. Ado’ expressed amazement at AI’s capability, saying, “At first listen, I thought—this AI voice sounds exactly like me!”

Recognizing Hualien and Taitung’s rich music scene, the Ministry of Culture chose Hualien Cultural and Creative Industries Park as the site for the Hualien AI Music Lab. Since its soft opening in January, the Lab has welcomed nearly 10,000 visitors from Taiwan and abroad, generating over 6,000 AI-assisted songs.

The Lab’s first-floor AI Experience Exhibition is open to the public for free, offering a QR Code-guided tour through the Music Hallway, where visitors can create AI-powered songs in the Creative Otherworld, experience their AI-generated music videos played at the “Pick & Play Warehouse”, and capture their “Superstar Moment” at the Starlight Avenue photo zone.

For more information on upcoming lectures, courses, and performances, visit the official Hualien AI Music Lab website: https://hmusic.moc.gov.tw.

/by
,

National Dong Hwa University and Taiwan AI Labs Launch NDHU-GPT: Transforming Campus Life with AI

[January 23, 2025] National Dong Hwa University (NDHU) and Taiwan AI Labs announced the successful development of the AI campus assistant “NDHU-GPT” at a joint press conference held on January 23, 2025. This cutting-edge AI solution has significantly enhanced administrative efficiency by saving nearly 1,000 hours of repetitive Q&A tasks for faculty and staff. Moreover, it acts as a comprehensive and intelligent assistant, providing students with precise and unbiased answers to a wide range of campus-related inquiries.

Ethan Tu, Founder of Taiwan AI Labs, highlighted that this collaboration harnesses the locally developed FedGPT platform, a state-of-the-art large language model tailored to meet the needs of over 10,000 NDHU faculty and students. During its training process, FedGPT integrated knowledge from more than 5,000 documents, encompassing critical information such as admissions policies, course registration procedures, and health insurance applications for international students. Additionally, it covers everyday details like dining options on campus or the availability of dormitory supplies. By consolidating essential information from various departments and incorporating internal institutional knowledge, NDHU-GPT delivers an unparalleled level of accuracy and efficiency in campus support services.

Photo Caption: Ethan Tu , Founder of Taiwan AI Labs (left), and Hui-Mi Hsu , President of National Dong Hwa University (right), pose for a photo.

Revolutionizing Campus Operations with AI

Since its initial deployment in 2024, NDHU-GPT has demonstrated tangible benefits, including:

  • Saving 1,000 Administrative Hours: The assistant’s knowledge base spans over 10 million words sourced from 5,000 documents, including admissions guidelines, course registration policies, and campus regulations. By automating responses to common queries, NDHU-GPT eliminates the need for time-intensive staff training, saving nearly four months of manual work.
  • 24/7 Real-Time Support: Operating around the clock, NDHU-GPT delivers instant answers to students and staff. During the freshmen enrollment period, the AI assistant handled over 1,000 inquiries, reducing wait times and resolving issues more efficiently.
  • Enhanced Support for International Students: With support for both Traditional Chinese and English, NDHU-GPT bridges language barriers, helping international students navigate complex procedures such as health insurance applications and scholarship inquiries. This feature alone has saved approximately 125 working days per semester for administrative staff.

A Secure and Reliable AI Assistant

Unlike open-source models, NDHU-GPT is built on the FedGPT platform, which prioritizes security and accuracy. It operates independently of public clouds, ensuring data privacy and preventing the integration of irrelevant or unverified information.

“NDHU-GPT functions as a well-trained, reliable employee—one that never sleeps and always provides consistent, accurate answers,” said Ethan Tu.

The model’s tailored design ensures it delivers context-specific responses. For instance, when asked about campus shuttle schedules, NDHU-GPT provides precise routes and timings rather than vague links to external websites. The assistant also handles sensitive or personal queries, such as where to go on a date or how to cope with exam stress, offering thoughtful, supportive answers drawn from official and culturally appropriate sources.

To ensure ethical AI deployment, NDHU-GPT is equipped with safeguards to prevent biased or inappropriate content, reinforcing NDHU’s commitment to responsible AI use.

At the event, a live demonstration showcased NDHU-GPT responding to common questions from students, including:

“What’s the best place to visit in Hualien during the weekend?”

“I’m feeling homesick—what should I do?”

“How can I improve my grades without feeling overwhelmed?”

The AI assistant provided helpful and empathetic guidance, reflecting its role as a supportive, trusted companion for students navigating campus life.

Photo Caption: The AI campus assistant NDHU-GPT operates 24/7, answering student queries about campus life challenges.

NDHU’s Commitment to Innovation and AI Literacy

President Hsu highlighted NDHU’s recent achievements, including its ranking as 86th globally in the 2024 World University Rankings for Innovation and 11th in “Generative AI Applications.” Hsu emphasized that NDHU-GPT represents the university’s vision of integrating cutting-edge technology into education while preparing students for an AI-driven future.

“Through this partnership with Taiwan AI Labs, we are not only improving operational efficiency but also fostering a new generation of AI-literate talent. Our students will graduate with the skills and confidence to collaborate with AI responsibly,” said Hsu.

Looking ahead, NDHU plans to expand its AI initiatives by introducing AI-based courses, developing virtual teaching assistants, and collaborating with industry partners to further advance the smart campus vision.

About FedGPT

FedGPT (https://yating.tw/zh-hant/fedgpt-landing) is an advanced large-language model developed by Taiwan AI Labs. Built using over 100 billion Traditional Chinese tokens, it emphasizes trustworthy AI principles, leveraging internal knowledge bases and federated learning for validated and reliable outputs.

About Taiwan AI Labs

Taiwan AI Labs (https://ailabs.tw) pioneers trustworthy AI solutions, integrating generative AI across healthcare, federated learning, virtual avatars, and other applications. The lab’s research spans AI music, speech recognition, and media analysis, earning recognition in leading international journals. By uniting talent, resources, and industries, Taiwan AI Labs drives AI innovation from Taiwan to the world.

Media Contact

Ellie Chen

PR, Taiwan AI Labs

Email: wanhsin.chen@ailabs.tw

/by
, ,

2025 Taiwan AI Industry Annual Conference: Ethan Tu Outlines Taiwan’s Path to Global Leadership

[January 22, 2024] 

The generative AI wave is driving innovation and business transformation worldwide, establishing itself as a core trend in 2025. Today, Taiwan AI Labs hosted the 2025 Taiwan AI Industry Annual Conference, featuring key industry leaders such as Ethan Tu, Founder of Taiwan AI Labs; Liang-Gee Chen, former Minister of the Ministry of Science and Technology and Honorary Professor at National Taiwan University; Ray-Jade Chen, Chairman of Taipei Medical University; and Bo-Yen Shen, Chairman of the Chinese Software Industry Association. Discussions covered topics ranging from the progression of generative AI to Artificial General Intelligence and the emerging concept of Regenerative AI. The conference aimed to equip Taiwanese enterprises with strategic insights to excel in the global AI landscape.

Photo Caption： The “Master Dialogue” panel at the 2025 Taiwan AI Industry Annual Conference featured moderator Chao-Hwei Hwang, Chief Content Officer of Taiwan AI Labs, along with panelists Liang-Gee Chen, former Minister of the Ministry of Science and Technology and Honorary Professor at National Taiwan University; Ray-Jade Chen, Chairman of Taipei Medical University; Bo-Yen Shen, Chairman of the Chinese Software Industry Association; and Ethan Tu, Founder of Taiwan AI Labs.

2025: The Era of Global AGI Competition – Opportunities and Challenges for Taiwan

The global AI market is forecasted to reach $243.7 billion by 2025, according to Statista. At the conference, experts outlined the evolution of AI development from GenAI to AGI, emphasizing three key trends: domain expertise, issue-awareness, and self-improvement.

Future AI models are expected to move away from being generalized tools, shifting toward specialized GPTs tailored for specific industries. These advanced GPTs will be designed to solve precise problems, adapt to dynamic environments, and continuously optimize performance.

Despite its potential, generative AI faces significant challenges, particularly its high development costs. For instance, Microsoft’s OpenAI reportedly spent $41 million to $78 million on developing ChatGPT-4, while Meta employed over 16,000 H100 GPUs for its latest model training. Such investments highlight the steep financial and technical barriers for AI adoption. Additionally, the EU’s Artificial Intelligence Act, which came into effect in August 2024, introduces stricter compliance and transparency requirements, including content labeling obligations. Non-compliance risks severe penalties, such as fines reaching 7% of global revenue.

In this competitive global environment, Taiwan’s challenge—and opportunity—lies in developing sustainable, trustworthy, and responsible AI solutions that align with these international regulatory trends, paving the way for Taiwan to strengthen its global presence.

Photo Caption：Ethan Tu highlighted that providing sustainable AI solutions is Taiwan’s opportunity to leverage its soft power and make a mark on the global stage.

Strategies for Success:  Vision for AI Development in Taiwan

Ethan Tu identified three critical strategies for positioning Taiwan as a global leader in AI innovation:

  1. Building Affordable and Energy-Sustainable GPTs The increasing adoption of AI has significantly raised energy consumption, with data centers accounting for up to 12% of national electricity use in the U.S., according to Deloitte. Taiwan’s geopolitical advantages and robust network infrastructure make it an ideal host for AI data centers, but its limited energy resources necessitate careful management.
    To address this, AI models are evolving into task-specific, energy-efficient tools tailored to business needs. Taiwan AI Labs’ FedGPT, for example, is a financial AI assistant designed to serve banks by offering precise, domain-specific expertise. Its professional version, FedGPT Pro, has demonstrated superior performance in financial knowledge tests while maintaining a smaller model size, reducing energy requirements.
    FedGPT employs federated data governance, allowing businesses to train custom AI models internally without the extensive use of external cloud resources. This approach minimizes data migration and energy consumption, offering a sustainable solution for enterprises.
  2. Proactively Adapting to Advanced Regulations As AI governance and data privacy regulations tighten globally, compliance with standards such as the EU’s GDPR and AI Act is increasingly crucial. Tu emphasized the importance of adopting “trustworthy technology” and addressing biases and explainability in AI systems.
    Early compliance with these regulations not only mitigates risks but also accelerates access to international markets. Taiwan AI Labs’ federated learning approach allows organizations to develop models in a secure, privacy-preserving manner, ensuring alignment with evolving global standards.
  3. Leading with a National-Level Federated GPT Platform Taiwan’s expertise in high-speed computing positions it as a key player in training advanced AI models. Taiwan AI Labs has pioneered a national-level federated data governance platform, launching FedGPT services to enable organizations to optimize AI models while safeguarding data privacy.
    This federated learning approach strengthens data security and enhances model accuracy, particularly for sectors handling sensitive information. By championing this innovative framework, Taiwan can contribute to shaping a more responsible and sustainable AI future.

Taiwan AI Labs 2025 Vision: Advancing FedGPT Capabilities Across Industries

In 2024, Taiwan AI Labs collaborated with key industries to deploy FedGPT across healthcare, finance, education, and multimedia entertainment sectors. By addressing core industry needs, FedGPT has accelerated digital transformation and fostered innovation.

Looking ahead to 2025, FedGPT will evolve into a versatile AI Agent, incorporating advanced capabilities such as vision-language models (VLM) for image, video, and audio analysis, and automatic speech recognition. With features like facial recognition and object detection, FedGPT will further expand its applications to meet diverse enterprise requirements, reinforcing Taiwan’s role as a leader in the global AI landscape.

Photo Caption：Taiwan AI Labs partners with multiple industry collaborators to introduce FedGPT into various sectors, including healthcare, finance, education, and multimedia entertainment.

Key Features of FedGPT

  1. AutoKB: Quickly transforms an organization’s internal knowledge into a GPT-powered knowledge base, enabling precise and professional responses.
  2. AutoML: Builds task-specific models on top of the foundational model to address unique business requirements.
  3. Infodemic: Launching in 2025, this feature analyzes and summarizes real-time global issues, distinguishing between genuine public opinion and online misinformation. Industries such as finance can leverage this capability to produce accurate KYB (Know Your Business) reports, enhancing decision-making and market insights.
  4. AutoCopilot: Simulates business scenarios and integrates with internal APIs and knowledge bases to automate complex internal processes. Fully deployed, it accelerates new employee development by learning from internal experts, driving organizational efficiency and performance.

Excellence Center Program: Empowering Sustainable AI Talent in Taiwan

At the 2025 Taiwan AI Industry Annual Conference, Taiwan AI Labs unveiled the “Excellence Center Program,” designed to advance AI applications and talent development in Taiwan’s universities through cutting-edge technology and resource integration.

The program’s inaugural phase brings together six esteemed scholars, including:

  • Professor Hung-Yi Lee, Department of Electrical Engineering, National Taiwan University
  • Professor Yun-Nung Chen, Department of Computer Science, National Taiwan University
  • Professor Chia-Yu Hsu, Department of Industrial Management, National Taiwan University of Science and Technology
  • Associate Professor Min-Fan Lee, Institute of Automation and Control
  • Professor Rua-Huan Tsaih, Department of Information Management, National Chengchi University
  • Professor Chao-Tung Yang, Department of Computer Science, Tunghai University

Program Highlights

  1. Driving Academic Applications and Enhancing Research Competitiveness
    By leveraging Taiwan’s proprietary large language model, FedGPT, the program equips universities with cutting-edge tools to improve research efficiency in areas such as literature summarization, data analysis, and data cleaning. This initiative aims to elevate Taiwan’s academic presence and global competitiveness in AI research.
  2. Developing AI Research Ethics Guidelines
    In collaboration with academic institutions, the program will establish AI research ethics guidelines to uphold principles of trustworthiness and responsibility. These guidelines will safeguard data privacy and intellectual property, ensuring that AI applications remain ethical and secure.
  3. Cultivating World-Class AI Talent
    The program offers secure platforms and technical support to enhance AI skills for students and faculty. Additionally, it aims to attract international talent to Taiwan while encouraging top local talent to return, strengthening Taiwan’s AI ecosystem and boosting its global competitiveness.

About FedGPT

FedGPT (https://yating.tw/zh-hant/fedgpt-landing) is an advanced large-language model developed by Taiwan AI Labs. Built using over 100 billion Traditional Chinese tokens, it emphasizes trustworthy AI principles, leveraging internal knowledge bases and federated learning for validated and reliable outputs.

About Taiwan AI Labs

Taiwan AI Labs (https://ailabs.tw) pioneers trustworthy AI solutions, integrating generative AI across healthcare, federated learning, virtual avatars, and other applications. The lab’s research spans AI music, speech recognition, and media analysis, earning recognition in leading international journals. By uniting talent, resources, and industries, Taiwan AI Labs drives AI innovation from Taiwan to the world.

Media Contact

Ellie Chen

PR, Taiwan AI Labs

Email: wanhsin.chen@ailabs.tw

/by
,

AI “Mentor” Helps Nurses Clock Out on Time, Garnering a 90% Satisfaction Rate: Taiwan AI Labs and Hualien Tzu Chi Hospital Showcase Breakthroughs at 2024 Healthcare+ Expo

December 6, 2024 — The 2024 Healthcare+ Expo, Asia Pacific’s premier AI healthcare event, highlights the collaboration between Hualien Tzu Chi Hospital and Taiwan AI Labs, unveiling three cutting-edge innovations: the FedGPT-powered system for nursing shift handovers and care summaries, AI-generated admission and discharge records, and a virtual health educator fluent in Taroko language. These innovations drew significant attention at the expo.

President of Hualien Tzu Chi Hospital Lin Hsin-Jung noted, “Even our international colleagues in Tzu Chi are amazed by Taiwan’s advanced progress.” As the only medical center in eastern Taiwan, the hospital handles over 800,000 patient visits annually, supported by a team of more than 1,000 nurses. Since integrating AI-powered solutions from Taiwan AI Labs, FedGPT has demonstrated an accuracy rate of over 80% in creating care records and shift handovers, significantly reducing the workload on nursing staff and earning a satisfaction score of 4.7 out of 5.

Photo Caption: Hualien Tzu Chi Hospital President Lin Hsin-Jung (left), Taiwan AI Labs Founder Ethan Tu (center), and Hualien Tzu Chi Hospital Vice President Wu Pin-An (right) pose for a group photo.

FedGPT Streamlines Nursing Tasks: AI “Mentor” Supports New Nurses in Timely Shift Handovers

Hualien Tzu Chi Hospital’s Director of Nursing, Chung Hui-Chun, emphasized the challenges that new nurses face, particularly during shift handovers, where detailed documentation of patient conditions, vital signs, and care priorities for the next eight hours is essential. Traditionally, this process relied on senior nurses dictating key points while junior nurses manually recorded the information—a labor-intensive process prone to errors and omissions, extending working hours and impacting care quality.

Photo Caption: Taiwan AI Labs Founder Ethan Tu (left) and Hualien Tzu Chi Hospital Director of Nursing and Virtual Health Educator Prototype Lead, Chung Hui-Chun (right), pose for a photo.

With FedGPT, nurses can now record conversations and generate comprehensive shift handover notes at the press of a button. These notes include actionable reminders, such as medication schedules or upcoming check-ups, presented in a structured format. Acting as a “virtual mentor,” FedGPT not only compiles reports but also ensures essential tasks are not overlooked.

Photo Caption:With a single press of the recording button, FedGPT can analyze doctor-patient conversations to summarize medical conditions and generate shift handover records.

Chung observed that many junior nurses, who completed their internships online during the COVID-19 pandemic, required three to four months of hands-on guidance to become proficient. Some took nearly a year. With FedGPT’s support, nurses adapt more quickly, enabling experienced staff to focus on critical tasks and improving overall team efficiency. This advancement aligns with the hospital’s vision of a “new normal” for nursing: seamless handovers and on-time clock-outs.

FedGPT: Pioneering Medical GPT Technology with Precision and Security

Developed by Taiwan AI Labs, FedGPT is a globally unique generative AI model integrating automatic speech recognition (ASR), text-to-speech (TTS), virtual avatars, and a federated learning platform. Tailored for healthcare, the model was trained on over 15 billion biomedical tokens, ensuring domain-specific expertise.

As AI adoption grows in healthcare, safeguarding sensitive medical data and mitigating algorithmic bias are critical challenges. Ethan Tu, founder of Taiwan AI Labs, highlighted the organization’s commitment to privacy protection through federated learning—a model-sharing approach that avoids data sharing. This ensures hospitals can train robust medical models without compromising patient confidentiality.

Despite using fewer computational resources, FedGPT achieves exceptional performance. In benchmarking tests using the Open Medical-LLM dataset, FedGPT Pro demonstrated superior medical knowledge comprehension among 14-billion parameter models, rivaling larger-scale models like GPT-4. Its efficient design also supports ESG goals by minimizing energy consumption.

Since deploying AI-assisted nursing solutions six months ago, Hualien Tzu Chi Hospital has continuously refined its systems based on frontline feedback. Notable achievements include the virtual health educator’s ability to provide guidance in 18 languages, including Taroko—a global first. Patients’ suggestions, such as incorporating friendlier avatar expressions, have also led to thoughtful enhancements, demonstrating the hospital’s dedication to integrating warmth into healthcare technology.

Director Chung Hui-Chun expressed optimism about AI’s transformative potential: “By alleviating repetitive tasks, empowering patients, and equipping nurses with advanced tools, AI can redefine healthcare delivery. Supporting our nurses today ensures a healthier future for all.”

About FedGPT

FedGPT (https://yating.tw/zh-hant/fedgpt-landing) is an advanced large-language model developed by Taiwan AI Labs. Built using over 100 billion Traditional Chinese tokens, it emphasizes trustworthy AI principles, leveraging internal knowledge bases and federated learning for validated and reliable outputs.

About Taiwan AI Labs

Taiwan AI Labs (https://ailabs.tw) pioneers trustworthy AI solutions, integrating generative AI across healthcare, federated learning, virtual avatars, and other applications. The lab’s research spans AI music, speech recognition, and media analysis, earning recognition in leading international journals. By uniting talent, resources, and industries, Taiwan AI Labs drives AI innovation from Taiwan to the world.

Media Contact

Ellie Chen

PR, Taiwan AI Labs

Email: wanhsin.chen@ailabs.tw

/by
,

Taiwan AI Labs Releases One Year Cross-Platform Analysis on Foreign Interference in U.S. Elections

[December 4, 2024] — Taiwan AI Labs has released a year-long observational report today, shedding light on extensive foreign influence operations targeting the U.S. presidential election. The report presents substantial evidence indicating interference by foreign actors, including Russia and China, aiming to propagate the narrative that the United States will fall under authoritarian control regardless of the election outcome. These tactics exacerbate polarization within the U.S. political landscape, intensifying societal division and posing a threat to democratic integrity.

Utilizing Taiwan AI Labs’ cognitive security tool, Infodemic, the report analyzed data from October 2023 to November 2024 across major social platforms such as Facebook, YouTube, X, Weibo, and TikTok. It examined 383 key narrative battlegrounds and scrutinized 330 reports from state-affiliated media outlets in Russia and China.

Key Narratives Exploited: Anti-Semitism, Foreign Election Interference, Immigration, Judicial Cases, and Gun Violence Fuel U.S. Decline Narratives

The report identifies major themes exploited by coordinated troll groups, including anti-Semitic conspiracies, immigration issues, judicial controversies, and narrative of American decline. Specific focus areas included:

From October 2023 to June 2024, Five major troll groups’ early operations focused on Trump’s immunity issues and Biden’s health concerns. After July’s “Trump Assassination Attempt,” narrative marked a significant shift, where trolls expanded from Supreme Court criticism to questioning government legitimacy and gun rights issues.

The Trump-Harris campaign launch in late July triggered another surge in coordinated activities. Operations expanded to include conspiracy theories about Jewish control of American politics through the “Tim Walz VP Candidate” narrative, while using Musk as a focal point during the “Trump-Elon Live Interview” to suggest threats to social stability.

Foreign interference intensified in September through two main strategies: Troll groups on YouTube, Facebook, and TikTok portraying both candidates as “colluding with Russia,” while troll groups on X focused on dismissing Russian interference concerns while highlighting the “Hunter Biden laptop” controversy. Troll groups on Weibo uniquely employed both strategies through sarcastic amplification of Putin’s alleged support for both candidates.

October operations shifted toward domestic issues – candidates’ health, immigration, disaster response, and wealth inequality. Following reports of Chinese state-backed hacking targeting campaign communications, post-election activities on Weibo intensified significantly. These operations promoted narratives about inevitable U.S. social division and decline, while questioning the integrity of leadership transition through narratives about Biden’s sudden “mental recovery” and Harris’s electoral defeat.

Troll Groups Amplify State Media Narratives to Undermine Trust in Democratic Alliances

The report highlights three main strategies used by troll groups to align with state-affiliated media narratives:

1.The narratives and strategies of troll groups frequently echoed Russian or Chinese state media.  

2.One of the main strategies was to shape candidates as colluding with Russia, spread the impression that the U.S. would be controlled by authoritarian regimes regardless of the election outcome and polarize American politics, further dividing society. They accused both Trump and Harris of colluding with Russia, criticized U.S. economic and foreign policies, and exploited topics like border and immigration issues to erode trust in both parties.

3.Besides U.S. domestic issues, troll groups also targeted foreign topics especially on NATO, EU tariffs, and U.S.-China relations. The narratives undermine the relationship of Western alliances, attack the policies against China, and describe China as a victim of U.S. aggression.

Taiwan AI Labs emphasizes the importance of addressing complex, cross-platform information warfare in the digital age. The findings highlight the urgent need for robust cognitive security solutions to safeguard democratic processes globally.

Full Report Link:One Year Cross-Platform Analysis of Troll Activities and Foreign Influence in 2024 U.S. Presidential Election

For further information or media inquiries, please contact:

Ellie Chen

Public Relations, Taiwan AI Labs

Email: wanhsin.chen@ailabs.tw

About Infodemic

Infodemic is a cognitive security tool developed by Taiwan AI Labs that monitors information in real time across social platforms. Utilizing AI technology, it aims to uncover information manipulation and coordinated actions, prevent the spread of malicious content, and identify the strategies employed by coordinated account groups.

The Infodemic platform leverages FedGPT to expose information manipulation, reveal coordinated behaviors, trace the spread of harmful content, and recognize the narrative tactics used by coordinated groups. Through its cognitive security scanning system, organizations can verify any type of content—whether URLs, images, videos, or even sentences. By simply inputting data and initiating a scan, the system determines if the content is organic or the product of malicious actors exploiting platform algorithms to manipulate public perception.

Upon detecting coordinated activity, the platform provides detailed insights into malicious networks, hidden agendas, and bot activities operating across the internet and social media. With comprehensive infographics, organizations can easily navigate the complex landscape of information warfare and assess its impact. Additionally, the system explores potential connections between these actors and foreign entities.

About Taiwan AI Labs

Taiwan AI Labs is dedicated to developing internationally trusted AI solutions through open algorithms, open-source code, and federated learning. Leveraging Taiwan’s talent, semiconductor industry strengths, and extensive healthcare data, the organization drives advancements in smart healthcare, human-machine interfaces, and generative AI.

Under the principles of trustworthy AI, Taiwan AI Labs effectively integrates generative AI into solutions such as smart healthcare, federated learning, AI virtual anchors, future scenario-generating studios, and news analysis platforms. In the human-machine interface field, the lab develops AI music generation, speech recognition, and semantic understanding technologies. These innovations address globally critical issues like disinformation and fake news, mapping the patterns of information manipulation. Our research has sparked international interest, receiving recognition and publication in leading scientific journals.

/by