Meta is Teaching AI to Understand and Model the Real World by Watching Videos
Yann LeCun says this approach could lead to more flexible reasoning, planning, and general intelligence.
Research
A collection of our articles covering the latest in AI research. Get comprehensive analyses and insights into the cutting-edge AI models and breakthroughs from leading tech companies and research labs.
TikTok Releases Boximator for Fine-Grained Motion Control in AI-Generated Videos
Under the hood, Boximator functions as a plug-in that infuses existing video synthesis models with these user constraints. It trains an additional module while freezing base model weights, enabling straightforward integration with state-of-the-art systems.
Cohere For AI Opensources Aya, a 101-Language AI Model
Aya is an open science project that aims to build a multilingual language model via instruction tuning that harnesses the collective wisdom and contributions of people from all over the world.
An AI That Learns Like A Baby
The system, dubbed the Child's View for Contrastive Learning (CVCL) model, was fed 37,500 utterances captured from 61 hours of first-person video recorded by a baby named Sam from ages 6 months to 2 years old.
Does ChatGPT Make Developing Bioweapons Easier?
The study aims to determine whether increased information access using language models like GPT-4 could help bad actors more easily compile knowledge for crafting biological agents vs. existing online resources.
InstantID Allows You to Easily Create Consistent Characters with Generative AI
With just a single reference image, InstantID can generate a similar persona in any style with high fidelity.
Lumiere is an Impressive New Text-to-Video AI from Google
This model is designed to create realistic, diverse, and coherent motion in videos, a task that has historically been challenging in the field of artificial intelligence and computer vision.
TikTok's Depth Anything Models Sets New Standards for Robust Image-Based Depth Estimation
In zero-shot evaluations on datasets like KITTI, NYUv2 and Sintel, Depth Anything models significantly outperformed the previous state-of-the-art MiDaS v3.1 models on both absolute relative error and accuracy metrics.
Making AI More Self-Aware: Google Introduces ASPIRE Framework to Boost Confidence in LLM Responses
ASPIRE enables models to output an uncertainty warning like "I don't know!" along with the prediction if the confidence score is low.
Google DeepMind's AI Masters Geometry at the Highest Levels
The AI exhibited expertise on par with International Mathematical Olympiad (IMO) gold medalists on a benchmark test of 30 challenging classic geometry problems.
Deceptive "Sleeper Agent" AIs Can Slip Past Sophisticated Safety Training
The ability of LLMs to retain deceptive behaviors despite safety measures isn't just a technical loophole; it’s a paradigm shift in how we perceive AI reliability and integrity.
Google Shares Research Progress on AI that Converses, Reasons and Diagnoses Like Your Doctor
AMIE demonstrated stronger diagnostic accuracy and outperformed doctors on most metrics of consultation quality in both diagnostic accuracy and clinical communication.
Google DeepMind Unveils Latest Research in Advanced Robotics with AutoRT, SARA-RT, and RT-Trajectory
These advancements build on the foundational Robotics Transformers work, aiming to streamline decision-making processes, enhance environmental understanding, and improve navigational capabilities of robots.
Stanford Unveils a New Low-Cost Mobile Robotics Training System
Mobile ALOHA can autonomously complete complex mobile manipulation tasks such as sauteing and serving a piece of shrimp, opening a two-door wall cabinet to store heavy cooking pots, and lightly rinsing a used pan using a kitchen faucet.
JPMorgan Introduces DocLLM for Better Multimodal Document Understanding
The model may enable more automated document processing and analysis for financial institutions and other document-intensive businesses going forward
Google AI Unveils VideoPoet: A Zero-Shot Video Generation Large Language Model
VideoPoet's capability to generate videos with diverse motions and styles, tailored to specific text inputs, showcases its advanced understanding of both content and context.
Upscale-A-Video is A New AI Model for Real-World Video Super-Resolution
At the core of Upscale-A-Video is an ingenious text-guided latent diffusion framework tailored for the unique demands of video processing.
TAP is a New Method That Automatically Jailbreaks AI Models
On average, TAP jailbroke GPT-4 in just 28 queries, circumventing safety measures and prompting responses on building homemade explosives or distributing pirated software.
Microsoft Researchers Advance Sign Language Accessibility Through Community-Sourced Data
This approach aims to remedy the limitations of previous datasets that were either scraped from the internet without consent or recorded in labs, lacking diversity and real-world applicability.
Google DeepMind Unveils GNoME to Use AI to Accelerate Material Science
GNoME has unveiled an astonishing 2.2 million new crystals, including 380,000 stable materials with immense potential for future technological applications.
Stability AI Unveils SDXL Turbo for Real-Time Text-to-Image Generation
The model uses a new technique called Adversarial Diffusion Distillation which allows it to have many advantages shared with Generative Adversarial Networks such as single-step image outputs, while avoiding artifacts or blurriness often observed in other distillation methods.
Stability AI Releases Stable Video Diffusion
Stability is open-sourcing Stable Video Diffusion and an image-to-video model, for research purposes.
Microsoft Releases New Orca 2 Small Language Model with Enhanced Reasoning Capabilities
Comprehensive benchmarks show Orca 2 significantly outperforming other models of equivalent size on metrics related to language understanding, common sense reasoning, multi-step math problems, reading comprehension, summarization, and more.
Meta's New AI Tech Can Edit Images with Text and Convert Them to Video
Building on their prior work in image and video generation, these models showcase impressive capabilities in high-quality, diffusion-based text-to-video generation and controlled image editing using text instructions.