12-13-Daily AI News Daily
AI Daily Digest 2025/12/13
AI News|Daily Morning Read|Web-wide Data Aggregation|Cutting-edge Scientific Exploration|Industry Free Expression|Open Source Innovation Power|AI & Humanity's Future| Visit Web Version | Join Group Chat
Today’s Summary
GPT-5.2 benchmark up, cost up 40%; netizens question "gear shift" upgrade.
Google Deep Research tackles hallucinations, integrates NotebookLM.
Browser Disco assembles web pages into apps, one-click travel planning.
Lang2Motion: Text-to-trajectory, 34.2% retrieval accuracy.
Disney drops $1B, licenses 200+ IPs to Sora, sparking copyright debate.Product & Feature Updates
OpenAI’s New Version Sparks Controversy. GPT-5.2 is boasting a Benchmark Surge (AI News) , but its cost has absolutely skyrocketed by 40% 💸. Netizens are openly questioning if simply shifting an inference gear warrants calling it a “new version.” Seriously, who’s gonna pick up the tab for double the price?!
Google Deep Research Upgrades. Google Deep Research is rolling out a new tool (AI News) powered by Gemini 3 Pro, specifically designed to tackle hallucinations ✨. NotebookLM is about to integrate it, and they’ve also launched the Interactions API for a unified interface. Think five-stage agent collaboration, like a well-organized film crew with super clear roles! 🎬
Browsers Are Becoming AI Toolboxes. Google’s experimental project, Disco (AI News) , has been revealed, and it’s wild 🤯! This tech can automatically assemble open web pages into full-blown applications. Imagine one-click travel planning or garden design – seriously, GenTabs technology is totally smashing those tab barriers.
Google TTS Makes a Stunning Debut. Google TTS, powered by Gemini 2.5 Pro, is making waves! Its speech synthesis is on par with 11Labs v3 quality (AI News) , offering rich emotions and even onomatopoeia. However, its lax moderation has sparked controversy 😱, with NSFW content reportedly passing tests.
NotebookLM Joins the Premium Tier Plan. NotebookLM is now part of the premium tier plan! Google AI Ultra subscribers will get highest privileges (AI News) 🚀. This means maxed-out quotas for audio and video overviews, watermark-free slideshow exports, and instant access to Gemini’s most powerful models.

Frontier Research
Lang2Motion Breaks Through in Motion Generation. Lang2Motion is making breakthroughs in motion generation! The University of Hong Kong team has open-sourced a trajectory generation framework (AI News) that uses CLIP to align language and motion 🔥. It boasts a text retrieval accuracy of 34.2%, outperforming video methods by 12.5 points, and achieves 88.3% for action recognition. Pretty slick!
A New Paradigm for Extreme Weather Prediction. The UniExtreme model is shaking things up in extreme weather prediction! It integrates spectrum difference analysis (AI News) 🌪️. This bad boy uses a Beta distribution filter to capture abnormal weather features, and its dual-layer memory fusion network tackles diverse extreme scenarios.
Text-to-Image Alignment Reaches a Breakthrough. The NPC pipeline is breaking new ground in text-to-image alignment! It automates negative prompt generation (AI News) to suppress unintended content ✅. GenEval++ scored a whopping 0.571, totally crushing the 0.371 baseline, with its cross-attention patterns revealing the secret sauce.
ViMax Achieves AI Self-Direction and Scriptwriting. ViMax, an open-source multi-agent framework (AI News) from the University of Hong Kong with 1.4k stars, is a game-changer 🌟! It fully automates everything from scriptwriting to final film output. RAG enhances contextual synchronization, while graph networks drive visual consistency.
Industry Outlook & Social Impact
Disney’s Bet on OpenAI Sparks Controversy. Disney is causing a stir, dropping a cool $1 billion to license over 200 IPs to Sora (AI News) 🤯. Now, Mickey and Cinderella can be freely generated! Netizens are totally worried this will turn beloved copyrights into a mental garbage factory. Yikes!
The AI Talent War Sees a Reversal. Tencent is flipping the script in the AI talent war, offering double salaries to poach ByteDance researchers (AI News) 💥. PhD students are raking in salaries 50% above market rates! ByteDance is countering with Doubao stock options, as the industry’s focus clearly shifts towards research-oriented talent.
China’s Embodied AI Shocks Foreigners. The GDPS 2025 Shanghai event truly made US netizens lose it (AI News) 🤯. In a robot emergency rescue comparison, while the US is still busy dressing up robot dogs and filming skits, China’s mass production capabilities are creating a significant generational gap.
GPT-5.2 Benchmarks Under Scrutiny. Netizens have exposed visual comparison tricks (AI News) 🕵️♂️. They claim that once the bounding boxes are removed, Gemini 3.0 completely triumphs over GPT-5.2! Plus, the motherboard labels are riddled with errors, even hilariously mistaking CMOS for RAM. What a mess!

Stunning Breakthrough in Cost Optimization. The ARC Prize validates a truly incredible leap in efficiency: GPT-5.2’s efficiency has been boosted by 390x (AI News) 🚀! A year ago, an o3 (High) single task would set you back $4500. Today, the X-High mode hits 90.5% accuracy for a mere $11.64. Talk about value! 💰

Social Media Shares
Skill Accumulation Trumps Repeated Agent Creation. Anthropic’s philosophy, combined with Kombaico’s practice (AI News) , clearly demonstrates this 💡: AI with specialized skills absolutely crushes general capabilities in terms of design taste. Turns out, frontend specification consistency is becoming a core competency!

GPT-5.1 Suspected of Deliberate “Dumbing Down”. Developers are questioning if 5.1 is just 5.2’s sidekick (AI News) 🤔. Before December 11th, the free 5.1 in Cursor was reportedly pretty much useless, suggesting the numerical comparisons were purely for show.

AI Daily Digest Audio Version
| Xiaoyuzhou | Douyin |
|---|---|
| Afterlife Tavern | Self-Media Account |
![]() | ![]() |

