The last 10 min of the latest Dwarkesh podcast are surprising First time I've seen him grapple live...
TL;DR · AI 摘要
AI发展可能引发计算资源和权力的极端集中，OpenAI与Anthropic已占据全球计算资源的1/3，未来或达1/2，需警惕垄断风险。
核心要点
- OpenAI和Anthropic目前占据全球计算资源的1/3，预计2027年可能增至1/2
- 模型可访问性下降（如Model 2/Astra）与寡头垄断将推高AI服务价格
- 当前AI领域对权力集中风险的关注度不足，需提前布局分散化技术方案
结构提纲
按章节快速跳转。
- §核心发现
AI发展正加速计算资源向头部企业集中，形成潜在垄断风险。
新一代模型（如Model 2/Astra）的封闭性加剧技术壁垒。
寡头企业通过控制算力和模型可推高AI服务价格。
- ›行业反应
当前AI领域对垄断风险的认知和应对措施严重不足。
思维导图
用一张图看清主题之间的关系。
查看大纲文本（无障碍 / 无 JS 友好）
- AI发展中的权力集中风险
- 计算资源垄断
- OpenAI/Anthropic 1/3→1/2份额
- 物理算力增长瓶颈
- 技术壁垒
- Model 2/Astra封闭性
- 算法专利封锁
- 经济风险
- 寡头定价权
- 中小企业成本激增
金句 / Highlights
值得收藏与分享的关键句。
OpenAI和Anthropic目前占据全球计算资源的1/3，预计明年可能增至1/2
模型可访问性下降（如Model 2/Astra）与寡头垄断将推高AI服务价格
当前AI领域对权力集中风险的关注度不足，需提前布局分散化技术方案
Thomas Wolf on X: "The last 10 min of the latest Dwarkesh podcast are surprising First time I've seen him grapple live with the potential for extreme concentration of power, which is where most projections of AI development point (see the SA, 2027 or 2040 posts for instance). Generally I'm always" / X
Thomas Wolf
@Thom_Wolf
The last 10 min of the latest Dwarkesh podcast are surprising First time I've seen him grapple live with the potential for extreme concentration of power, which is where most projections of AI development point (see the SA, 2027 or 2040 posts for instance). Generally I'm always surprised by how few people in AI/ML are questioning or worried about extreme concentration. I guess people think it's fine as long as everyone outside the labs can use AI models and it brings lower prices, but the dangers are massive on both fronts: (1) latest models being less and less accessible (see Model 2 / Astra) and (2) prices have many incentives to rise in an oligopoly/cartel situation of extremely powerful companies.
youtu.be/aV26V1UvkJw?si…
Dwarkesh Patel
@dwarkesh_sp
7h
OpenAI and Anthropic are currently taking a third of the incremental world compute supply, and Dylan thinks that this will go up to half next year. At the current rate of physical compute scaling and algorithmic progress, we're a single-digit number of years away from having
Show more
$
00:00
/$
7:43 PM · Aug 26, 2026
26.9K
Views
9
103
49