Charts
DataOn-chain
VIP
Market Cap
API
Rankings
CoinOSNew
CoinClaw🦞
Language
  • 简体中文
  • 繁体中文
  • English
Leader in global market data applications, committed to providing valuable information more efficiently.

Features

  • Real-time Data
  • Special Features
  • AI Grid

Services

  • News
  • Open Data(API)
  • Institutional Services

Downloads

  • Desktop
  • Android
  • iOS

Contact Us

  • Chat Room
  • Business Email
  • Official Email
  • Official Verification

Join Community

  • Telegram
  • Twitter
  • Discord

© Copyright 2013-2026. All rights reserved.

简体繁體English
|Legacy
BTCBTC
💲71814.00
+
2.89%
ETHETH
💲2101.58
+
2.76%
SOLSOL
💲89.21
+
3.84%
USDCUSDC
💲0.9999
-
0.01%
XRPXRP
💲1.43
+
3.62%
DOGEDOGE
💲0.09726
+
4.81%

a16z
a16z|10月 13, 2025 14:08
Columbia CS Professor: Why LLMs Can’t Discover New Science From GPT-1 to GPT-5, LLMs have made tremendous progress in modeling human language. But can they go beyond that to make new discoveries and move the needle on scientific progress? Distinguished Columbia computer science professor Vishal Misra argues against. LLMs compress the extremely complex world into Bayesian manifolds, and while confidence is high on the manifold, LLMs hallucinate when reasoning outside of their training data. A true AGI wouldn’t just be able to reason across larger and larger manifolds, but create new ones entirely. 0:00 Intro 0:32 LLMs and humans reason through manifolds 4:15 Token prediction, entropy & confidence 10:20 Vishal’s background 14:10 Inventing RAG 17:30 The question of progress plateauing 21:00 The Matrix Model 28:10 Why LLMs can’t recursively self-improve 34:02 Defining AGI 38:25 Future architectures 42:00 Modeling vs prompt engineering 47:20 What would prove AGI has arrived? 50:01 Closing thoughts @vishalmisra @martin_casado @eriktorenberg(a16z)
+2
Mentioned
|
APP
Windows
Mac
Share To

X

Telegram

Facebook

Reddit

CopyLink

|
APP
Windows
Mac
Share To

X

Telegram

Facebook

Reddit

CopyLink

Timeline

11月 01, 13:36gpt-5-pro model has intelligence reduction
10月 29, 01:38Cysic has done a great job in terms of implementation.
10月 27, 08:05MiniMax is open source and has launched the MiniMax M2 model
10月 27, 02:17Cysic Network is a Layer-1 blockchain built on Cosmos CDK.
10月 26, 14:57ComputeFi's new paradigm becomes the mainstream architecture for computing power
10月 23, 14:00Amjad Masad and Marc Andreessen discuss AI agents
10月 20, 12:36Sentient has 4 papers accepted at NeurIPS
10月 19, 01:06The accuracy issue of large models in new information processing
10月 17, 09:15Mira builds AI consensus layer and trust layer
10月 14, 17:05Is Artificial Intelligence Slowing Down? Asking the Wrong Question

HotFlash

|
APP
Windows
Mac
Share To

X

Telegram

Facebook

Reddit

CopyLink

APP
Windows
Mac

X

Telegram

Facebook

Reddit

CopyLink

Hot Reads