MULTIMODAL AI

→ View original post on X — @jeffdean

Gemma Multimodal Models Progress on American Sign Language Translation

By

@jeffdean

–

21 May 2025 19h16

Here's one good sign: there's been some really nice progress on adapting our open Gemma multimodal models to translate American Sign Language to English. https://t.co/FBUvHoKMfO
— Jeff Dean (@JeffDean) 21 mai 2025

Here's one good sign: there's been some really nice progress on adapting our open Gemma multimodal models to translate American Sign Language to English.

→ View original post on X — @marek_rosa

Google Veo 3 Generates Stunning Space Engineers Video Content

By

@marek_rosa

–

21 May 2025 17h58

Space Engineers as imagined by Google Veo 3 👍🙂 https://t.co/X0FMLOCnth
— Marek Rosa | European🇪🇺 | South African🇿🇦 (@marek_rosa) 21 mai 2025

Space Engineers as imagined by Google Veo 3

LLM excels in MMMU and USAMO

By

–

21 May 2025 16h51

84% on MMMU and leading the USAMO? This isn’t just an LLM, this is a research assistant with a calculator brain.

Google’s AI Agent Mode for Complex Tasks

By

–

21 May 2025 16h40

8/ Agentic AI: Google's AI can now do things for you.

The Gemini app is getting an "Agent Mode" that can handle complex tasks like apartment hunting – filtering listings, scheduling tours, and comparing options, all autonomously. pic.twitter.com/Dgu3UhXktW
— God of Prompt (@godofprompt) 21 mai 2025

8/ Agentic AI: Google's AI can now do things for you. The Gemini app is getting an "Agent Mode" that can handle complex tasks like apartment hunting – filtering listings, scheduling tours, and comparing options, all autonomously.

Gemini Live Adds Project Astra Features

By

–

21 May 2025 16h40

6/ Project Astra:

Your universal AI assistant is here.

Gemini Live now incorporates Project Astra's camera and screen-sharing capabilities, allowing the AI to understand your surroundings. pic.twitter.com/Z6kRXddr9m
— God of Prompt (@godofprompt) 21 mai 2025

6/ Project Astra: Your universal AI assistant is here. Gemini Live now incorporates Project Astra's camera and screen-sharing capabilities, allowing the AI to understand your surroundings.

Incredible Image Generation Breakthrough

By

–

21 May 2025 16h39

2/ Imagen 4: Prepare to be amazed by visual fidelity. Out of this world detail: Image generation just took a giant leap forward, with incredible realism, fine details, and improved typography. Generate images in various aspect ratios and up to 2K resolution.

Google’s AI Nuclear Bomb: Key Updates

By

–

21 May 2025 16h38

🚨 BREAKING: Google just dropped an AI NUCLEAR BOMB.

Imagen 4, Veo 3, Flow, AI mode in search, Gemini 2.5 Pro Deep Think, Project Astra, and Agent mode.

Here's everything you need to know: pic.twitter.com/syDXsSDdrp
— God of Prompt (@godofprompt) 21 mai 2025

BREAKING: Google just dropped an AI NUCLEAR BOMB. Imagen 4, Veo 3, Flow, AI mode in search, Gemini 2.5 Pro Deep Think, Project Astra, and Agent mode. Here's everything you need to know:

→ View original post on X — @christinelu

AI Facial Expression Limitations and Sarcasm Detection Challenges

By

@christinelu

–

21 May 2025 16h32

I think it’s also because she can’t move her face. So even if she was trying to be playfully sarcastic as some say, it’s hard to tell.

→ View original post on X — @tunguz

YouTube Data Access Gives Google Massive AI Video Model Advantage

By

@tunguz

–

21 May 2025 16h19

So it turns out that having direct access to all of YouTube footage gives you a huge leg up in terms of training AI video models. Who would have guessed.