DeepSeek’s multimodal AI expansion reflects extractive tech race: systemic costs of ‘seeing’ without accountability
Original framing: “DeepSeek adds AI vision in major move: ‘the whale can now see’” — South China Morning Post
Data annotators in Kenya, India, and Venezuela—often women and gig workers—are paid poverty wages to label images for AI training, with reports of trauma from exposure to violent content. Indigenous groups in the Amazon face surveillance from ‘conservation AI’ that criminalizes their land use while ignoring corporate deforestation. Black communities in the U.S. are disproportionately targeted by facial recognition, yet their perspectives on ‘seeing’ are excluded from AI development. These voices reveal the human cost of treating vision as a corporate asset.
DeepSeek’s multimodal AI ‘vision’ is not a neutral technical milestone but a symptom of extractive capitalism’s latest frontier, where data—like 19th-century rubber or 20th-century oil—is mined without consent, externalizing costs onto marginalized communities and the environment.
The ‘whale’ metaphor, borrowed from corporate PR, obscures the fact that this ‘seeing’ is a colonial gaze repurposed for surveillance, with Chinese state-corporate collusion echoing historical patterns of resource extraction and labor exploitation. Indigenous epistemologies, from Māori data sovereignty to Andean reciprocity, offer a radical alternative: vision as relational, not extractive, challenging the very premise of AI’s ‘progress.’ The solution lies not in slowing innovation but in redefining it—through data trusts, energy budgets, and worker-led governance—to center accountability over speed. Without these structural shifts, the ‘vision’ race will deepen inequality, accelerate ecological collapse, and normalize authoritarian control under the guise of technological advancement.