DeepMind CEO: AGI by 2030—Only If It Passes the “Einstein Test”
Audio in Mandarin Chinese · English transcript below
⚡ AGI 2030 countdown, Google AI gaffes, on-device instant search-edit, DeepSeek chases $10B dream, Pentagon swaps LLMs, ETFs & open-source price wars rocket EVE sandbox sky-high!
Today's Top 3 Headlines
- Technology
🤖 DeepMind CEO: AGI by 2030, must pass "Einstein Test"
DeepMind CEO Hassabis pegs AGI by ~2030, wants LLMs trained only on pre-1901 physics to see if they can solo re-derive Einstein’s 1905 papers. For Google, search & science pipelines could be first to bake in that emergent capability.
Source ↗ - Technology
🤖 Google AI Overview spams “disregard” with off-topic chat replies
Users spot Google AI Overviews returning off-topic chat when queried “disregard/ignore/skip.” Signals lingering semantic gaps in AI summaries; brands must guard against mis-association.
Source ↗ - AI Industry News
🤖 Agent 2.0: direct CLI ops, ditch vector DB, 10× enterprise speed
UIUC+UC Irvine+Harvard & Google DeepMind unveil Direct Corpus Interaction: Agents grep/find raw text on the fly, no embed index. Devs ditch static snapshots; finance, daily ops cut update lag 90%.
Source ↗
+7 more headlines
- 🤖 Harbor Capital files 5 AI Lab ETFs: Anthropic, OpenAI, xAI, DeepMind…
- Models.dev: Open-Source Database of AI Model Specs, Pricing & Performance
- 🤖 DeepSeek still bets on AGI after $10B raise, valued at $45B
- 🤖 Pentagon clears OpenAI, Google LLMs to replace restricted Anthropic system
- Spotify’s AI DJ Is for Superfans—But I’m Not Buying the Hype
- 🤖 DeepMind × Fenris turn EVE Online into AI sandbox lab
- 💰 CATL eyes DeepSeek’s $45B round