diffuse.one
diffuse.one · 2,761 words · saved by 2 readers
andrew white's blog.
diffuse.one diffuse.one/ virtual_cell designation: D1-009 author: andrew white updated date: September 26, 2025 human cell atoms: 10 14 10^{14} 1 0 14 FLOPs for human cell: 2 × 10 38 2 \times 10^{38} 2 × 1 0 38 expected simulation year: 274 power requirement 200 TW abstract: The people want a virtual cell. So let's simulate an entire human cell atom-by-atom for 24 hours with molecular dynamics. I estimate it will take 2 × 10 38 2\times10^{38} 2 × 1 0 38 FLOPs and 200 TW of electricity. Based on historic frontier simulations, we should be able to reach this milestone in 2074. I based on this an
saved by
related reading
- Scaling Laws That Extrapolate 300× Past the Fitopenathena.ai
- Fermi estimate of future training runsdanieldewey.net
- Ushering in the Thermodynamic Future | Extropicextropic.ai
- Levers for Biological Progress - by Niko McCartyasimov.press
- My picture of the present in AI — LessWronglesswrong.com
- Why didn't we get GPT-2 in 2005?dynomight.net
- The Moon Should Be a Computerpalladiummag.com
- AI Timelines — LessWronglesswrong.com
- Can AI scaling continue through 2030? | Epoch AIepoch.ai
- Why compute might get 10x more expensive in coming yearsdwarkesh.com
- Dario Amodei — "We are near the end of the exponential"dwarkesh.com
- Dylan Patel — Deep dive on the 3 big bottlenecks to scaling AI computedwarkesh.com