Davy K.

#AIExpert #Entrepreneur. Retweets & likes are not endorsements. Msc IT (AI & Software Engg) @CarnegieMellon | Ph.D. | board: @c4ir_rw.

[António Guterres @antonioguterres]

On the International Day of Reflection on the 1994 Genocide Against the Tutsi in Rwanda, we mourn the victims & pay tribute to the survivors.

We must learn from past failures & protect the living — by rejecting hatred, inflammatory rhetoric & incitement to violence.

[Government of Rwanda @RwandaGov]

Today we commemorate the 32nd anniversary of the 1994 Genocide against the Tutsi and begin the official commemoration week.

Remember - Unite - Renew
#Kwibuka32

[AK @_akhaliq]

model: https://t.co/ouihs3nh9V

[Presidency | Rwanda @UrugwiroVillage]

This morning, President Kagame opens the Land Forces Commanders Symposium, convening Land Forces Commanders and military leaders from over 30 African countries and beyond to discuss strategies to enhance peace and security on the continent. Follow live: https://t.co/dpxiJwdgc6

[AK @_akhaliq]

Ditto

Scaling Instruction-Based Video Editing with a High-Quality Synthetic Dataset

[AF Post @AFpost]

The US Army has sworn in senior tech executives from Palantir, Meta, and OpenAI as lieutenant colonels in a new reserve unit, where they will serve as advisors to steer the implementation of AI in military strategy.

Follow: @AFpost

[AK @_akhaliq]

VMem

Consistent Interactive Video Scene Generation with Surfel-Indexed View Memory

[Tanishq Mathew Abraham, Ph.D. @iScienceLuvr]

Zero-Shot Vision Encoder Grafting via LLM Surrogates

"We construct small “surrogate models” that share the same embedding space and representation language as the large target LLM by directly inheriting its shallow layers. Vision encoders trained on the surrogate can then be directly transferred to the larger model, a process we call zero-shot grafting – when plugged directly into the full-size target LLM, the grafted pair surpasses the encoder-surrogate pair and, on some benchmarks, even performs on par with full decoder training with the target LLM. Furthermore, our surrogate training approach reduces overall VLM training costs by ∼45% when using Llama-70B as the decoder."

[Tanishq Mathew Abraham, Ph.D. @iScienceLuvr]

Towards Scalable Language-Image Pre-training for 3D Medical Imaging

"In this study, we introduce Hierarchical attention for Language-Image Pre-training (HLIP), a scalable pre-training framework for 3D medical imaging. HLIP adopts a lightweight hierarchical attention mechanism inspired by the natural hierarchy of radiology data: slice, scan, and study"

[Google AI @GoogleAI]

We’ve been amazed by the excitement around Veo 3, which brings in a new era for video generation. It's state of the art in text-to-video, image-to-video, text-to-audio+video generation, and realistic physics.

Veo 3 generates videos with audio – from background noises to dialogue, excelling at accurate lip syncing all from a single text prompt.

And now, we are expanding access to Veo 3 in 71 new countries. And Pro subscribers can now try it in both the Gemini app and in Flow, our new AI-filmmaking tool. Can’t wait to see what you create. ✨

[AK @_akhaliq]

AdaptThink

Reasoning Models Can Learn When to Think

[Tanishq Mathew Abraham, Ph.D. @iScienceLuvr]

Unlocking Non-Invasive Brain-to-Text

"Here, we present the first non-invasive B2T result that significantly exceeds these critical baselines, raising BLEU by 1.4-2.6x over prior work. This result is driven by three contributions: (1) we extend recent word-classification models with LLM-based rescoring, transforming single-word predictors into closed-vocabulary B2T systems; (2) we introduce a predictive in-filling approach to handle out-of-vocabulary (OOV) words, substantially expanding the effective vocabulary; and (3) we demonstrate, for the first time, how to scale non-invasive B2T models across datasets, unlocking deep learning at scale and improving accuracy by 2.1-2.3x."

[International Security Conference on Africa @ISC_Africa]

H.E. Paul Kagame, President of the Republic of Rwanda’s full remarks at the inaugural International security on Africa.

🔗https://t.co/l8whE6CuMR

[Watcher.Guru @WatcherGuru]

JUST IN: 🇺🇸🇸🇦 Nvidia $NVDA announces 'strategic partnership' with 'Humain' to build AI factories in Saudi Arabia.

[DeepLearning.AI @DeepLearningAI]

Researchers introduced Mender, a recommendation system that uses Llama 3 to infer customer preferences from text like product reviews and descriptions.

Instead of relying directly on raw text, which can contain noise, Mender extracts explicit preferences, such as a liking for bold colors, and uses them to predict future purchases. Tested on Amazon and Steam datasets, it outperformed traditional systems like TIGER, especially in categories with highly variable customer tastes.

Read our summary of the paper in The Batch: https://t.co/XsXtQnv4CT

[The Hacker News @TheHackersNews]

🚨 $1.9 BILLION Crypto Swap Site Seized!

Germany’s BKA has shut down eXch[.]cx, a crypto exchange used by North Korean hackers & darknet actors.

🔍 €34M in crypto seized
📁 8 TB of data confiscated
🌐 Ran on clearnet + dark web

Read full story ➝ https://t.co/b03Ef3LGYr

[CNBC Tech @CNBCtech]

Alphabet shares sink after Apple’s Cue says AI will replace search engines https://t.co/zMihjp8xqK

[kabi @kakakbibibi]

🌐 Introducing WebThinker, a powerful open-sourced deep research agent!

🧐 WebThinker autonomously conducts in-depth web searches and generates comprehensive research reports.

🔑 We release the technical details in our paper！

💡HF paper: https://t.co/ZigN3wVUhS

[Rwanda Broadcasting Agency (RBA) @rbarwanda]

WATCH

President Kagame and the First Lady Jeannette Kagame have arrived at Kigali Genocide Memorial for the National Remembrance Ceremony.

#Kwibuka31 #RBANews

[Presidency | Rwanda @UrugwiroVillage]

This morning at the Kigali Genocide Memorial, President Kagame and First Lady Jeannette Kagame mark the beginning of the 31st commemoration of the 1994 Genocide Against the Tutsi. Follow live: https://t.co/NGtGw1QDAa