🚀 Google TPU 8 slams Nvidia, Cannes debuts AI film fest
Audio in Mandarin Chinese · English transcript below
⚡ Google drops $40B, TPU 8 dukes it out with Nvidia, DeepSeek V4 runs on your couch, AI pills hit humans, Cannes bans AI flicks, UK & US hoard power and fight piracy—this week’s subway ride feels like New Year’s Eve!
A central trend in the current AI domain is the full-scale race and deep strategic deployment by tech giants across both hardware and software ecosystems. Google released its TPU 8 chip, directly challenging Nvidia's dominance in the cloud compute market, while simultaneously making a massive investment in Anthropic, further solidifying its strategic advantage in the AI software domain. Viewed together, this not only signals an accelerating iteration of AI technology from foundational compute to advanced LLMs but also sparked attention and discussion, such as at the Cannes Film Festival, regarding AI-generated content, demonstrating AI's profound impact across various industries.
Today's Top 3 Headlines
- Miscellaneous
Google TPU 8 Slams Nvidia’s AI Dominance—Pune Mirror
Google unveils TPU 8-series AI chips to **slash training time & latency**, directly challenging Nvidia’s cloud dominance and signaling fierce compute rivalry.
Source ↗ - Technology
Cannes Debuts First AI Film Festival, Sparking Global Buzz
**Cannes** hosts first World AI Film Festival, screening AI-generated movies—**contrast** to official festival’s AI ban.
Source ↗ - Technology
🤖 DeepSeek V4 preview drops, tuned for Huawei Ascend chips
DeepSeek V4 preview ships Huawei Ascend kernel-level tuning, cutting inference latency 18%. Domestic devs can now run 175B LLMs without Nvidia, saving 30% hardware cost.
Source ↗
+7 more headlines
- ⚖️ US State Dept: DeepSeek AI "distillation" theft alert
- Anthropic: Google to Pour $40B into AI Startups
- 🤖 First AI drug enters human trials: DeepMind’s Isomorphic Labs leads
- 🤖 UK AI data centers may need 6 GW by 2030—two agencies’ forecasts differ 10×
- AI Weekly Rewind: April 26, 2025 – Patrick McGuinness on How AI Is Rewriting Everything
- 🤖 Mainframe→AI: 60 yrs on, clients still want results
- 🤖 DeepSeek V4 preview drops, near-top LLM runs locally