AI-powered
podcast player
Listen to all your favourite podcasts with AI-powered features
Preventing Rogue Training of Dangerous AI Models
To prevent the training of dangerous AI models in rogue settings, implement compute tracing by tracking the sales of GPUs from centralized sources like TSMC and ASML. By controlling access to powerful hardware like Nvidia's AI training chips A100 and H100, through compute governance, rogue actors can be hindered from acquiring the necessary resources to train such models.