New AI video models, insane humanoid robots, quantum chips, new drugs, Grok 3
Welcome to the AI Search newsletter. Here are the top highlights in AI this week.
Step-Video-T2V is a powerful text-to-video model with 30 billion parameters. It can generate videos up to 204 frames long and uses advanced techniques like deep compression Video-VAE and Direct Preference Optimization. Read more
Do you prefer to watch instead of read? Check out this video covering all the highlights in AI this week:
xAI has released Grok 3, which outperforms some of the best models across math, science, and coding benchmarks. The model features advanced reasoning and search capabilities, including "Think," "Big Brain," and "DeepSearch" modes for different types of problem-solving and information retrieval. Read more
Microsoft's Majorana 1 is a groundbreaking quantum chip that uses a new Topological Core architecture. It leverages topoconductors, a novel material that can control Majorana particles to create more reliable qubits. This innovation could lead to quantum computers capable of solving complex industrial and societal problems in the near future. Read more
We’re partnering with NVIDIA to giveaway a RTX 6000 Ada (48GB VRAM). Simply register and attend their upcoming GTC event, a premier conference covering a wide range of topics from AI to robotics to quantum computing. Free online sessions available. Enter the giveaway here
Google Research has developed an AI co-scientist to accelerate scientific breakthroughs. This AI assistant can help researchers by analyzing data, suggesting experiments, and even generating hypotheses. It aims to enhance human creativity and productivity in scientific research across various fields. Read more
Phantom is an AI model by Bytedance that can generate high-quality videos from text descriptions, or reference images. It uses a novel approach to improve the coherence and quality of generated videos. The model can create diverse and realistic videos across various domains and styles. Read more
Pika releases Pikaswaps which allows users to easily swap elements in photos. Seamlessly replace objects in existing videos. The tool is designed to be user-friendly and accessible for both professionals and casual users. Read more
Figure has introduced Helix, a new AI model that allows humanoid robots to understand and perform complex tasks in homes. Helix combines visual perception, language understanding, and physical control, enabling robots to handle objects they've never seen before. Read more
Microsoft has created Muse, an AI model that can generate video game visuals and controller actions. This World and Human Action Model (WHAM) is the first of its kind and can produce game content or simulate gameplay. Microsoft is making the model's weights and sample data open-source for others to use and build upon. Read more
A good photo on your Linkedin or business profile makes a huge difference. You could do a physical photoshoot, which costs you over $200 and hours posing awkwardly at a camera. Or, with AI Portrait, just upload one photo, and get a portfolio of 50 professional photos in minutes. Save time and money - try it today!
Sakana AI has developed AI CUDA Engineer, a tool for automatically optimizing CUDA kernels. This framework can discover and improve CUDA code without human intervention, potentially speeding up GPU-based computations. It's the first comprehensive system of its kind for CUDA optimization. Read more
Mistral has released a new language model called Mistral Saba, designed for Middle Eastern and South Asian languages. The 24-billion parameter model is particularly good at handling South Indian languages like Tamil. It was trained on carefully selected datasets to ensure high-quality performance across multiple languages from these regions. Read more
OpenAI has introduced SWE-Lancer, a new benchmark to evaluate AI models on real-world freelance software engineering tasks. The benchmark consists of over 1,400 tasks from Upwork, with a total value of $1 million USD in actual payouts. Read more
Dynamic Concepts is a new AI by Snap that can understand and generate videos using existing videos as references. It uses a novel approach to capture the relationships between objects and their movements over time. Read more