OpenAI's Whisper is another case study in Colonisation

Whisper is a speech recognition model by OpenAI trained on 680,000 hours of audio. It can transcribe and translate in multiple languages. Despite its groundbreaking capabilities, Whisper didn't receive much attention due to the focus on generative AI models. The use of indigenous data in Whisper raises concerns regarding data extraction and sovereignty. The post argues that open sourcing models like Whisper primarily benefits those already in the tech industry, neglecting indigenous and marginalized communities. The decision of whether to open source such models should be made by the communities from whom the data was collected.