❶ Andrej Karpathy, former Tesla AI director, has demonstrated that reproducing GPT-2 can be done in 24 hours for just $672 using a single 8XH100 node, significantly cheaper than the $100 million cost of training GPT-4. ❷ The project, llm.c, directly implements GPT training in C/CUDA, eliminating the need for complex environments and speeding up the process. ❸ Despite advancements, AI training costs are not decreasing; leading-edge models like GPT-4 still require substantial investment, raising concerns about the environmental impact due to high power consumption.
Related Articles
- NVIDIA to supply 64,000 new AI GPUs for OpenAI, Oracle's new Stargate AI supercomputer in Texas3 months ago
- OpenAI has run out of GPUs, says Sam Altman — GPT-4.5 rollout delayed due to lack of processing power3 months ago
- OpenAI to tape-out on TSMC 3nm this year4 months ago
- DeepSeek AI begins to steal customers away from ChatGPT4 months ago
- OpenAI CEO Sam Altman responds to DeepSeek AI hype amid widespread stock value crash4 months ago
- Trump announces 0.5trn investment in AI4 months ago
- OpenAI execs mused over Cerebras acquisition in 2017 — to mitigate predicted Nvidia supply woes7 months ago
- OpenAI gets one of the first engineering builds of NVIDIA's new Blackwell DGX B200 AI system8 months ago
- NVIDIA to join Apple, Microsoft funding round for OpenAI that values the AI startup at $100B+9 months ago
- How Hard Can It Be?about 13 hours ago