Meta's 2022 launch of a model covering 200 languages, including 55 African languages, triggered immediate investor withdrawals from regional NLP startups. "Facebook has solved it, so your little puny startup is not going to be able to do anything," investors told African developers, according to Timnit Gebru, former Google AI ethics researcher.
The pattern extends beyond Africa. OpenAI representatives allegedly tell international language tech companies they'll "put you out of business soon" while offering minimal compensation for non-English training data, Gebru claims. The approach concentrates AI development in US tech giants while marginalizing regional innovation.
Gebru calls the dominant paradigm "stealing data, killing the environment, exploiting labor." Her critique targets companies racing to build universal models that claim to handle every task and language from centralized infrastructure.
Cognitive scientist Abeba Birhane argues "AI for good" messaging deflects criticism. "It allows companies to say 'Look, we're doing something good! Everything about AI is not bad. And you can't criticize us,'" she told AI Now Institute.
Recent developments challenge the necessity of massive centralized models. China's DeepSeek V4 demonstrated that resource-constrained development can produce competitive results, questioning whether billion-dollar compute budgets are inevitable. Nvidia just committed $4 billion to photonics infrastructure for larger models.
Gebru and Birhane advocate for task-specific, resource-efficient models over universal systems. Specialized models for particular languages or applications often perform better while consuming fewer resources and supporting diverse development ecosystems across regions.
The stakes are existential for developers working on low-resource languages from Southeast Asia to Latin America. Each Big Tech model announcement potentially triggers funding withdrawals, even when universal models underperform specialized alternatives on specific tasks.
The debate highlights tensions between Silicon Valley's scaling strategies and distributed global innovation. Critics question whether environmental costs and market consolidation justify marginal capability gains from increasingly large models.