Chinese company trained GPT-4 rival with just 2,000 GPUs — 01.ai spent $3M compared to OpenAI's $80M to $100M

Submitted ⁨⁨5⁩ ⁨weeks⁩ ago⁩ by ⁨BrikoX@lemmy.zip⁩ to ⁨technology@lemmy.zip⁩

With fewer resources, you have to maximize their efficiency.

Comments

Sort:hotnew top

jonathan@lemmy.zip ⁨5⁩ ⁨weeks⁩ ago
I’d like to hear it expressed in terms watts rather than number of GPUs for an indeterminate amount of time.

source
khannie@lemmy.world ⁨5⁩ ⁨weeks⁩ ago
In computer science in the late 90’s we were told (repeatedly) that many of the best algorithms for CPU and RAM usage came from the USSR. Necessity being the mother etc.

source
- possiblylinux127@lemmy.zip ⁨5⁩ ⁨weeks⁩ ago
  The USSR was way behind in computing
  
  source
  - khannie@lemmy.world ⁨5⁩ ⁨weeks⁩ ago
    Yes, way behind and that forced them to come up with better algorithms.
    
    source