The Improbable AI Lab at MIT proposes a new multimodal framework, HiP, for detailed planning. HiP uses three different base models and outperformed comparable frameworks in manipulation tasks. It operates hierarchically and combines language, physical, and environmental intelligence to assist robots in tasks like stacking blocks and arranging items. The research shows the potential of using existing pre-trained models for robot decision-making.

At CES 2024, TI launched 3 new chips to boost ADAS & BMS innovation in electric vehicles. Automakers are shifting focus to software-defined vehicles with crucial roles played by software and computing technology. Texas Instruments offers cutting-edge semiconductor solutions for next-gen vehicles with chips like AWR2544 boosting automation in ADAS and DRV3946 & DRV3901 enhancing BMS efficiency. TI aims to lead in semiconductor innovation for the electric vehicle future.

Major AI companies have invested billions in generative AI. Apple partners with media giants for exclusive copyright access. OpenAI pays $1-5 million annually for news article rights. Apple delves into on-device ML development with open-source Ferret model and 3D avatar research. Apple may collaborate with OpenAI on smartphone design. Siri expected to integrate generative AI. Apple's AI-related work in progress. Apple's partnership with news industry shows potential for growth.

Microsoft researchers introduce a cost-effective technique using Language Learning Model (LLM) to train custom embedding models, enhancing Retrieval-Augmented Generation (RAG) applications. They achieve state-of-the-art results surpassing popular embedding models in benchmark tests.

Baichuan Intelligence introduces Baichuan-NPC, a character model with contextual dialogue understanding. It offers a "Character Creation Platform + Enhanced Knowledge Base Search" solution for efficient character customization in gaming. Baichuan-NPC excels in dialogue, consistency, and role-playing.

At DingTalk 7.5 conference, AI Assistant launched for users and businesses to customize AI assistants. Aim to create 10m AI assistants in 3 years. Features include perception, memory, planning, action, and integration with third-party apps. Marketplace to facilitate transactions. Other upgrades include a Personal Edition with AI tools, upgraded Collaboration Tab, custom capabilities, and "Knock Knock" feature. Partnerships with Microsoft China, Xiaobing, TAL Education announced. Entrepreneurship

Mistral AI releases Mixtral 8x7B model, outperforming LLaMA 2 70B and GPT-3.5. It excels in MMLU test and surpasses Claude 2.1, GPT-3.5, and Gemini Pro. Mixtral has 45 billion parameters but utilizes 12 billion for inference per token, offering enhanced performance and cost efficiency over GPT-3.5. It supports multiple languages and code writing, with an optimized Instruct version achieving comparable performance to GPT-3.5.

Dutch AI startup Whispp uses assistive speech technology to convert speech impairments into a person's natural voice. The company secured €750,000 in funding to expand globally and won the CES 2024 Innovation Award. The technology aims to create a more inclusive world by enhancing human communication and connection.

Dr. Natalia Stakhanova, a security researcher, emphasizes the importance of recognizing the long-time presence of AI technology in our daily lives. AI can enhance security defenses but also presents cybersecurity risks. Utilizing AI for defense purposes can help test and protect computer systems. Stakhanova advocates for educating young people on cybersecurity and believes in the importance of establishing regulations for the responsible use of AI.

Apple is catching up in AI with new tools for iPhone, generative AI in iOS 18, and "Ajax" language model. However, full AI integration expected by 2025. New products include Vision Pro headphones, iPad Pro, iPhone 16, Apple Watch X, AirPods, iPad mini, Mac mini, and MacBook Pro.

Origin Quantum Computing Technology (Hefei) Co., Ltd. officially launched "Origin Wukong," China's third-generation superconducting quantum computer with 72 "Wukong Chips." It features the advanced "Origin Tianji" system for automated testing, enhancing operational efficiency. Named after Sun Wukong, the computer signifies China’s progress in quantum technology.

AI is set to dominate CES with innovations such as glasses for the blind and moonwalker shoes. Automotive tech and health devices will also be in the spotlight. Sustainability will be a key focus, with eco-friendly products showcased.

In 2024, AI faces challenges beyond basic research, such as cultural diversity and combining speech and images. Leaders predict inclusive AI models, open-source advancements, and transformative impacts on cybersecurity. Data centers will undergo major transformations to support AI growth.

OPPO launched the world's first GPT large-scale model smartphone, the Find X7 series. Features include AI voice summary and AI advantages for different scenarios, catering to various age groups.

Apple announced the release of Vision Pro mixed reality headset on February 2, at $3,499. CEO Tim Cook called it "the most advanced consumer electronics device ever." Apple aims to boost growth in the metaverse field where Meta lags behind.

Getty Images and Nvidia partner on Generative AI for creating stock photos. iStock's new AI platform caters to individuals, learning from Getty's library. Pricing starts at $14.99 for 100 prompts. Legal guarantees include $10,000 compensation limit. Inpainting and Outpainting features coming soon.

During the GTC conference, NVIDIA launched BioNeMo Cloud, a generative AI service suite. The updated cloud API integrates models from internal, open-source, and partner sources for drug discovery workflows. Amgen plans to utilize generative AI for drug discovery with the help of a system called Freyja. Recursion's Phenom-Beta excels in image reconstruction tasks, while NVIDIA's MolMIM generates small molecules. Various companies are utilizing BioNeMo for biological, chemical, and genomic researc

At CES 2024, McAfee introduced "Project Mockingbird" AI tech to combat deepfake audio scams. It detects fake audio with over 90% accuracy, using advanced AI models. Despite challenges, this tool marks progress in fighting digital deception.

Google is developing Bard Advanced, a paid subscription with advanced language skills and custom robot creation. There may also be AI enhancements to improve Bard prompts and expand content options. No official release date for these features yet.

Google's DeepMind announces advancements to improve robot decision-making. AutoRT teaches robots real-world tasks using language and visual models. SARA-RT enhances transformer models for efficient task execution. RT-Trajectory improves robot motion understanding with visual cues.