POPA ReconAGI as seen by ๐๐๐ (Demis Hassabis, Sergey Brin) source
AGI as seen by ๐๐๐ (Demis Hassabis, Sergey Brin) source 1. AI performance can continue to improve with current scaling alone. However, to reach AGI, new algorithmic breakthroughs are needed along with expansion of models, data, and computing. 2. In the future, demand for data centers will be greater in the inference process than in model learning. This is because the more complex the problem, the more chips and power the AI uses as it explores and verifies multiple thought paths in parallel. 3. AGI, as defined by Hassabis, is not at the level of automating average human tasks. It refers toโฆ
The excerpt is quoted from a public Telegram channel, which owns its text.