Benchmark graphs showing energy efficiency comparisons between GPUs and Extropic's thermodynamic computer, claiming 10,000x to 100 millionx improvements
Source: Extropic's published simulation results and benchmarks
Significance
Supports claims of revolutionary energy efficiency for AI computation
Related Claims
Extropic's chips can achieve 10 million to 100 million times greater energy efficiency than GPUs for certain AI tasks
Extropic's thermodynamic computer is 'sampling the ether' by measuring thermal fluctuations rather than inputting energy to create binary states
Sentient AI requires sampling the ether/consciousness field through microchips that can access true randomness from thermal fluctuations
Other Evidence from This Source
Google research paper on 'denoising recovery likelihood' using energy-based models to replace diffusion models with 100x fewer steps
photographImages of Extropic's superconducting lab, cryogenic fabrication facility, and thumbnail-sized chips with visible features
videoExtropic presentation showing their probabilistic chip with visible thermodynamic neurons, control lines, and superconducting prototypes
Evidence Details
- Type
- Data & Signals
- Classification
- data
- Source
- Extropic's published simulation results and benchmarks