In 2020, AlphaFold AI system achieved unprecedented accuracy in predicting protein structures. The latest model expands to DNA, RNA, ligands, and PTM complexes, accelerating discoveries in biology and drug development. It enables precise predictions for key molecular categories, offering new insights into disease pathways and potential therapeutic targets. AlphaFold's capabilities can enhance drug design by predicting ligand structures more accurately than traditional methods. The model's scalab

Obsidian, an open-source multimodal AI model with 3 billion parameters, breaks barriers by fitting on a regular laptop. It offers AI processing for text and images, similar to GPT-4V, without the need for expensive high-end hardware. The model's compact size, based on the Capybara-3B architecture, opens new possibilities for AI accessibility on devices. The upcoming Mistral model, also open-source, will further expand multimodal AI capabilities.

AWS announced new base models on Amazon SageMaker JumpStart: Code Llama & Mistral 7B. Code Llama offers optimized code generation & Mistral 7B is a large language model with 7B parameters, excelling in natural language tasks. Both models can be deployed with a click on AWS.

AI voice technology startup ElevenLabs launched a new platform, Eleven Labs Enterprise, offering advanced voice solutions with enterprise-level quality, security, scalability. The platform generates high-quality audio output meeting ACX audiobook standards. It emphasizes end-to-end encryption, optional privacy mode, compliance with SOC2, GDPR standards. CEO Alex Serdiuk aims to enhance workflows in publishing, call centers, gaming. The platform caters to publishers, call centers, media companies

With climate change leading to more frequent catastrophic floods, Multiverse Computing and partners are using quantum technology for improved flood risk assessment and prediction. This innovative approach aims to address computational challenges and enhance accuracy in large-scale flood models, offering hope in combating climate change impacts.

A new framework combining AI and data fusion technology improves sustainable product inspection. It addresses the disconnect between product lifecycle and AI applications for better predictions. The focus is on incorporating business data for predictive models in industries like automotive aftermarket. The framework enhances management of product returns and decision-making, promoting sustainability.

In the realm of image generation, AI tools like DallE 3, Adobe Firefly, and Midjourney offer unique features revolutionizing image creation. Each platform caters to different needs, with Adobe Firefly integrating seamlessly with existing Adobe workflows, DallE 3 focusing on detail and ethics, and Midjourney providing diversification based on community feedback.

Imitation Learning is essential for training neural networks for complex tasks by observing human behavior. Challenges include visual variations, but a new framework, GROOT, focuses on 3D representations for improved decision-making and performance in real-world tasks.

Baichuan Intelligence released Baichuan2-192K, the world's longest context window model, processing 350,000 Chinese characters. It outperforms Claude2 in text generation, understanding, and question answering. Achieving SOTA in 7/10 evaluation sets. Opening API for law, media, finance industries.

Global release of Wondershare Filmora 13, featuring AI-driven creative assistant Copilot for intelligent video creation. Enhanced cloud services for efficient collaboration. API testing open to core partners for document analysis and programming tasks. Leader in AI-driven content generation.

Kunlun Wanwei open-sources Tiangong Skywork-13B series large language model with 13B parameters and Chinese corpus dataset with 150B tokens. Models outperform others in Chinese text creation benchmarks. Download links provided.

Google released MediaPipe Image Generator plugin for high-end Android devices, allowing image generation on-device with fine-tuning options using diffusion models and LoRA weights for specific concepts. New plugins simplify adjustment process for generated images.

Alibaba Cloud's Qianwen app, a large-scale model communication app, is now available on the Android market with features like creative copywriting, office assistance, learning support, and fun Q&A. Users can download the 40.95MB app for various functions and register with their phone number. An iOS version is pending.

AMD is prioritizing AI on CPUs for edge devices, leveraging Siena for efficient processing. GPUs may face latency and power challenges on edge devices. AMD's broad AI strategy includes Instinct MI300X launch to compete with NVIDIA's H100. Expanding focus to edge computing and workload optimization.

Engineers at Binghamton University developed a robotic guide dog for the visually impaired. It responds to leash pulling, offering cost-efficient, accurate assistance. Future plans include adding natural language functionality and intelligent disobedience. The team seeks feedback from the visually impaired community to enhance the technology's usability. The goal is to have robotic guide dogs in public spaces within a few years.

Amazon has tested humanoid robots for repetitive tasks to improve efficiency. These robots, like Digit, aid in moving and grabbing objects, complementing human workers. AI integration in operations and security, along with robotic advancements, demonstrate Amazon's commitment to enhancing workforce.

Researchers at MIT's ICoN Center found evidence supporting "self-supervised learning" in the brain. Models trained this way showed similarities to brain activity. Another study focused on grid cells, showing a self-supervised model capable of path integration.

Microsoft CEO sees Copilot as the new "Start" button for Windows users. The integration of AI hardware like NPU in Intel's Meteor Lake and Qualcomm's Elite X SoC is expected to bring new PC experiences.

Bilibili introduces AI video summarization feature in beta test. Bilibili Index model supports search, content recommendation, and more. Double 11 initiatives drive traffic to e-commerce platforms. Q2 financial report shows user growth and engagement. CEO expects daily active users to surpass 100 million in the third quarter.

At the 2023 China Computer Conference, Zhifu AI launched ChatGLM3, a third-gen multimodal model excelling in text, image, audio, code understanding. With improved MML, CEval, GSM8K, BBH metrics. Offers code generation, web search knowledge enhancement, and AgentTuning for intelligent planning. Mobile and in-car endpoint testing models introduced. "Zhifu Qingyan" leverages ChatGLM3 for user interaction via code, search enhancement, image-text understanding. Zhifu AI aims for international recogni