AI-powered
podcast player
Listen to all your favourite podcasts with AI-powered features
Guardrails for Advanced AI Systems
The future potential of AI systems to develop dangerous capabilities necessitates the implementation of safeguards. These safeguards are unlikely to be chip-based due to the general nature of chips; instead, they are anticipated to be system-based. Deploying AI systems will require substantial computing power both for training and inference, with a focus on assessing the computational needs for inference based on the system's scale and ease of access.