Alibaba Launches AI Smart Glasses, Driven by Human-in-the-Loop Work

Alibaba is making a significant stride into the smart glasses market with a new device powered by its proprietary artificial intelligence models, a move that forms part of the company’s broader $52.4 billion investment in AI and cloud computing. Named Quark AI Glasses, this device marks Alibaba’s inaugural venture into the wearables category and is slated for launch in China by the close of 2025.

The Quark AI Glasses will integrate Alibaba’s Qwen large language model (LLM) with its AI assistant, Quark. While Quark is already available as a popular app in China, this marks the first instance of Alibaba pairing the AI assistant with dedicated hardware to broaden its user reach. As one of China’s most active AI developers, the Hangzhou-based tech giant has been consistently rolling out models designed to rival systems from global leaders like OpenAI. By entering the smart glasses arena, Alibaba joins an expanding cohort of tech firms betting on wearables as the next pivotal computing platform, potentially complementing or even succeeding smartphones.

The Quark AI Glasses will enter a competitive landscape that already features Meta’s smart glasses, developed in collaboration with Ray-Ban, and a model launched by Xiaomi earlier this year. Alibaba’s offering promises a range of hands-free functionalities, including calling, music streaming, real-time language translation, meeting transcription, and a built-in camera. Leveraging Alibaba’s extensive ecosystem of services in China, the glasses will also enable users to access navigation, make payments via Alipay, compare prices on Taobao, and utilize other Alibaba-owned platforms for mapping and travel booking. While a compelling feature set has been outlined, Alibaba has yet to disclose the device’s price or detailed specifications.

The sophisticated capabilities of smart glasses like Alibaba’s hinge on advanced AI systems capable of recognizing images, interpreting context, and responding in natural language. These abilities are fundamentally reliant on vast quantities of labelled data—information meticulously reviewed and tagged by humans to train the AI. This process frequently employs “human-in-the-loop” (HITL) systems, where human input is crucial at key stages of AI training and testing.

To illuminate the practicalities of HITL, Henry Chen, co-founder of Sapien—a company specializing in managing large, distributed workforces for data labelling—offered insights into common misconceptions, the escalating demand for skilled contributors, and the influence of China’s AI growth on the industry. Chen clarified that HITL extends far beyond simple data labelling, encompassing complex decisions on edge cases, nuanced judgment calls, and continuous evaluation. He emphasized that “continuous feedback is what makes HITL work instead of one-off datasets.”

Another prevalent misconception, according to Chen, is that HITL work is inherently low-skilled. He countered this by noting that the proliferation of industry-specific AI has generated a significant demand for domain experts—such as doctors, lawyers, and scientists—to contribute their specialized knowledge. Sapien, for instance, collaborates with 1.8 million contributors across 110 countries. For intricate tasks like contextual understanding or visual recognition, maintaining data quality is paramount. Chen explained that Sapien ensures consistent results through peer validation, meticulous contributor reputation tracking, and carefully aligned incentives.

China’s AI sector is undergoing rapid expansion, with demand for data labelling quickly approaching levels seen in the United States. Despite China’s unique regulatory environment, Chen observed that the types of AI projects are increasingly mirroring those in other major global markets. Managing such a large and dispersed workforce, Sapien utilizes on-chain technology to ensure payment transparency and empower its community to weigh in on which projects are pursued. By operating without traditional offices, Sapien claims to circumvent certain workplace issues, focusing instead on rewarding contributors based on the value they deliver.

While automation is undoubtedly transforming data labelling, Chen firmly believes that humans will remain central to specific types of work. Tasks involving cultural nuance, sarcasm, rare diseases, niche languages, or complex sentiment will continue to require human review. He predicted that “humans will shift focus towards long-tail data and new vertical domains,” foreseeing a future where AI-assisted labelling handles routine cases while people tackle the most challenging ones. For sensitive projects, such as those involving the intellectual property of large corporations or international organizations, stringent controls are necessary. Chen detailed Sapien’s approach, which includes vetting and training enterprise contributors, implementing data minimisation and access controls, and adhering to client-mandated compliance frameworks like SOC 2 Type 2, GDPR, and HIPAA.

As AI models improve at learning from unlabelled data—a process known as self-supervised learning—some anticipate a reduced need for human labelling. However, Chen views this not as a disappearance but an evolution of the human contributor’s role. He stated, “We will evolve into a more specialized industry,” highlighting Sapien’s increasing involvement in evaluating synthetic data and model outputs. Future projects, he expects, will concentrate on curating unique “ground truth” datasets, assessing AI performance, and providing highly specialized domain expertise.

Alibaba’s foray into smart glasses underscores the pervasive integration of AI into everyday products. While the Quark AI Glasses may be one among many wearable devices by 2025, the synergy of Alibaba’s in-house language model, its existing service ecosystem, and hardware integration could position them distinctly for users in China. Crucially, products like these rely on a complex supply chain of human expertise, from the engineers crafting the models to the contributors refining the data that powers them. Companies like Sapien operate behind the scenes, ensuring AI systems receive the precise information needed to function with accuracy and responsibility. Whether manifested as smart glasses, virtual assistants, or yet-to-be-conceived devices, AI-driven hardware is emerging as a critical avenue for companies to deliver their services directly to consumers. For Alibaba, the Quark AI Glasses represent both a product launch and a clear statement about its vision for growth—a future where technology seamlessly combines software, hardware, and indispensable human intelligence.