Microsoft is collaborating with OpenAI to develop sound recognition AI capable of detecting natural disasters by analyzing environmental sounds. The newly patented technology processes sound signals through a neural network, providing alerts for events like earthquakes and home intrusions. This AI integration could enhance the capabilities of applications like Copilot and ChatGPT, particularly for Windows users.
Meta has announced the next generation of its AI accelerator chip. Its development focused on chip memory (128GB at 5nm) and throughput (11 TFLOPs at int8).
Google has made its most advanced generative AI model, Gemini 1.5 Pro, available in public preview on its Vertex AI platform. It offers a large context window of up to 1 million tokens.
DreamView introduces an innovative approach to creating 3D objects from text descriptions that allows for detailed customization from multiple viewpoints while ensuring the object remains consistent overall.
In a study examining persuasiveness, the Claude 3 Opus AI model was found to closely match human persuasiveness. This was determined through statistical tests and adjustments for multiple comparisons. Humans were slightly more persuasive, but not by a statistically significant margin, underscoring a trend where larger, more sophisticated models are becoming more convincing. Claude 3 Opus emerged as the top model in terms of persuasiveness. A control condition confirmed the study's methodological reliability by showing expectedly negligible persuasiveness for indisputable facts.
Policy-guided diffusion offers a new method for training agents in offline settings, creating synthetic trajectories that closely align with both behavior and target policies. This technique helps to generate more realistic training data, significantly improving the performance of offline reinforcement learning models.
Attorch is an attempt to rewrite portions of PyTorch’s nn module in Python and Triton. It is designed to be an easily hackable and performant neural network experimentation library. It is important to note that this would have been prohibitively expensive to write only a few years ago.
Amazon invested an additional $2.75 billion into AI startup Anthropic, signifying the tech giant's focus on competing with Microsoft's OpenAI-powered services via AWS. Despite this investment, Amazon's internal AGI team is ambitiously working to outpace Anthropic with its own AI model, Olympus. Amazon's strategic move underscores the importance of advanced AI in the Big Tech arena.
Hash3D introduces a novel approach to accelerate 3D generative modeling by utilizing a hashing mechanism that leverages feature-map redundancy across similar camera positions and diffusion time-steps.
Elon Musk's xAI has released Grok-1.5, an AI with enhanced math and coding skills that boasts a significant performance increase and competitive benchmark results against leading AI models like GPT-4. The updated model can now process much longer context windows, improving its memory capacity. Grok-1.5 is currently accessible to Premium+ users of X. X plans to expand availability to regular Premium subscribers.
a16z conducted a deep dive into web traffic data to rank the most popular generative AI web products based on monthly visits and uncover patterns on how consumers are actually using Generative AI.