We're sunsetting PodQuest on 2025-07-28. Thank you for your support!
Export Podcast Subscriptions
DeepSeek, Deep Research, and 2025 Predictions with Sarah and Elad
2025/2/7
No Priors: Artificial Intelligence | Technology | Startups
E
Elad
以播客主持人身份，多次与行业领袖进行深入对话，探讨技术行业的最新趋势和预测。
S
Sarah
个人财务专家，广播主持人和畅销书作者，通过“Baby Steps”计划帮助数百万人管理财务和摆脱债务。
Sarah: 我认为DeepSeek的出现打破了中美技术由美国主导的局面,表明中国模型也能快速赶上。虽然DeepSeek的成本问题被误解,总成本远不止600万美元,但即使成本是600万美元的数倍,进入这个领域的门槛也远低于数十亿美元,这让市场感到震惊。人们不只是想要原始的下一个词的概率流,后期的训练和人类反馈至关重要。
Elad: 我认为DeepSeek在某些方面非常重要,但其发展也符合预期趋势。人们对DeepSeek感兴趣的原因有三:它是最先进的中国模型,在推理等方面赶上了西方模型;它是开源的;它声称成本很低。最终模型的运行成本大致在500万到1000万美元之间,关键在于其背后投入了多少工作。英伟达股价下跌有点过头了,DeepSeek背后可能花费了数亿美元的计算资源。
Deep Dive
Discussion about DeepSeek, a state-of-the-art Chinese AI model. The conversation covers its capabilities, cost-effectiveness, and the implications for the AI market, including model commoditization and the narrative around US vs. China AI dominance.
- DeepSeek is a state-of-the-art open-source Chinese AI model.
- Claims of low training costs ($5.5 million) are discussed, with counterarguments suggesting the actual cost was much higher.
- The model's impact on NVIDIA stock and the broader AI market is analyzed.
- The discussion touches upon model commoditization and the value of being a frontier model provider.
Shownotes Transcript