OpenAI Sora’s biggest rival is here: China's Kuaishou AI-Powered Video Generator Can Create Stunning 1080p Clips Up to 2 Minutes Long
Anybody Can Prompt (ABCP)
ABCP: Making Generative AI easy for everyone. Join us, because Any Body Can Prompt!
Today's highlights:
?? AI Breakthroughs
Kuaishou Technology Unveils Kling Model, Elevates Text-to-Video Standards
? Kuaishou Technology unveils Kling, a new text-to-video model capable of producing two-minute videos in 1080p high-definition ?
? Kling significantly outpaces OpenAI's Sora by generating more extended, realistic videos and is already available on the Kuaiying app by invitation ?
? Unlike its predecessor models, Kling supports complex animations like realistic motion and aspect ratio variations, setting a new benchmark in text-to-video AI technology.
Meta Empowers Small Businesses with AI Agents on WhatsApp for Enhanced Customer Engagement
? Meta introduces AI agents in WhatsApp Business app, aiding small businesses in efficient customer service and operations management ?
? The AI agents can automate responses to common queries on WhatsApp, enhancing customer interactions by quickly providing required information ?
? Meta's AI systems allow targeted messaging via WhatsApp, optimizing ad relevance and potentially increasing return on investment for businesses.
Audio Platform 'Udio' Updates Include WAV Downloads, Mobile Enhancements, and New Generative Features
? Udio introduced WAV downloads and audio uploads for enhanced control in music creation, available for Standard and Pro plan subscribers ?
? New two-minute model and advanced controls, like prompt or lyrics strength and generation-quality slider, launched to fine-tune musical output ?
? Subscriptions for audio inpainting feature are now available, allowing users to seamlessly edit audio tracks for more polished outputs.
Qwen2 AI Model Series Launched with Enhanced Multilingual Support and Advanced Capabilities
? Qwen2 updates include pretrained and instruction-tuned models across five sizes, ranging from 0.5B to 72B parameters ?
? The new models showcase proficiency in 27 additional languages, expanding beyond English and Chinese ?
? Qwen2 models deliver state-of-the-art results in numerous benchmarks, particularly excelling in coding and mathematics tasks.
NotebookLM Expands Globally with Google Slides Integration and Enhanced Fact-Checking Features
? NotebookLM, upgraded with Gemini 1.5 Pro, now supports Google Slides, web URLs, and includes better fact-checking through inline citations. ;
? Google's NotebookLM has expanded globally, now available in over 200 countries, enhancing research and writing processes. ;
? Real-world applications of NotebookLM range from authoring books to creating hyperlocal newsletters, demonstrating its versatility across various fields. ;
领英推荐
?? AI Ethics
OpenAI Expands Insight into Voice Engine Functionality and Safety Measures
? OpenAI's Voice Engine utilizes a text-to-speech model that requires only a 15-second audio sample to generate human-like speech launched in late 2022 for internal testing. ;
? As of September 2023, Voice Engine powers ChatGPT’s Voice Mode, using real voices selected through a comprehensive process starting in May 2023. ;
? OpenAI introduced a TTS API in November 2023, featuring six preset voices created with professional voice actors, targeting developers for website integration. ;
OpenAI outlines its architecture that supports the secure training of frontier models.
? OpenAI details its secure architecture for training advanced AI on Azure, utilizing Kubernetes for orchestration and Azure Entra ID for identity management ?
? The architecture includes robust security measures such as role-based access control and sensitive data storage using key management services ?
? OpenAI emphasizes continued evolution and improvement of security features to protect intellectual property and support safe AI advancement.
??AI Academia
Researchers Develop 'Buffer of Thoughts' to Boost AI Language Model Performance
? Buffer of Thoughts (BoT) enhances the accuracy, efficiency, and robustness of large language models by introducing a thought-augmented reasoning approach ?
? BoT's novel meta-buffer stores and dynamically updates thought-templates to improve problem-solving across various tasks, demonstrating significant performance boosts on 10 reasoning-intensive tasks ?
? Compared to existing models, BoT achieves up to a 51% performance improvement on tasks like Checkmate-in-One, while using approximately 12% of the costs associated with multi-query prompting methods.
No Language Left Behind (NLLB)- an AI model created by researchers at Meta capable of translation between 200 languages — including low-resource languages.
? A new massively multilayered neural machine translation model leverages transfer learning, targeting 200 languages, including low-resource ones. ?
? Advanced techniques, including a Sparsely Gated Mixture of Experts architecture, were used, improving translation quality by 44% over previous models. ?
? The model's performance was rigorously tested over 40,000 translation directions with a proprietary benchmark and human evaluation metrics, establishing a new standard in NMT. ; Read more
About us: We are dedicated to reducing Generative AI anxiety among tech enthusiasts by providing timely, well-structured, and concise updates on the latest developments in Generative AI through our AI-driven news platform, ABCP - Anybody Can Prompt!
Join our growing community of over 20,000 readers and stay at the forefront of the Generative AI revolution.