Today:

Figure Raises $675M at $2.6B Valuation and Signs Collaboration Agreement with OpenAI

AI robotics company Figure secures $675M in Series B funding at a $2.6B valuation, backed by Microsoft, OpenAI, NVIDIA, Jeff Bezos, and others. Teaming up with OpenAI, Figure aims to advance AI models for humanoid robots, focusing on language processing capabilities. Leveraging Microsoft Azure, they plan to expedite robot deployment for various real-world applications. 

Figure, comprising experts from renowned companies like Boston Dynamics and Tesla, also partners with BMW for commercial robot integration in automotive production. This investment fuels their mission to revolutionize AI and robotics, addressing labor shortages and enhancing global supply chains.

PR NEWSWIRE

Microsoft’s Windows 11 Copilot gets smarter with new plugins and skills

Microsoft's Windows 11 is boosting its Copilot feature, offering new skills and plugins. Users can soon enjoy Copilot handling tasks like toggling settings and using services like OpenTable and Shopify. These updates hint at a future where Copilot automates complex PC tasks. 

The addition of AI features to default apps like Clipchamp and Photos further enhances user experience. With upcoming plugins for restaurant and travel reservations, Microsoft aims to make Copilot a versatile PC assistant. These enhancements signify a shift towards a more automated and efficient computing experience for Windows 11 users.

TECHCRUNCH

Stack Overflow and Google Cloud Announce Strategic Partnership to Bring Generative AI to Millions of Developers

Stack Overflow and Google Cloud join forces in a strategic move, blending Stack Overflow's developer hub with Google Cloud's AI prowess. The partnership aims to equip developers with enhanced AI tools integrated into both platforms. Google Cloud will fuse its Gemini for Google Cloud with Stack Overflow, offering developers AI-driven coding assistance. Additionally, Google Cloud Console will incorporate Stack Overflow's validated technical knowledge. 

The collaboration targets improved developer productivity and innovation across various industries. Stack Overflow will harness Google Cloud's AI services to refine community engagement and content curation. The integration is set to roll out in the first half of 2024, showcasing the synergy between the two tech giants.

STACK OVERFLOW

Microsoft tailors AI for finance teams in specialized 'Copilot' strategy

Microsoft unveiled a specialized AI tool called Copilot for Finance, catering to finance departments. It aids in data review, report generation, and task management. Pricing and availability details are yet to be disclosed. This initiative aligns with Microsoft's strategy to tailor software for various industries. 

The Copilot series extends beyond finance, with potential future versions for marketing and supply chain tasks. This move intensifies competition with Salesforce, which recently introduced its AI assistant, Einstein Copilot. The shift suggests evolving business models in the tech landscape.

REUTERS

Google expands BigQuery with Gemini, brings vector support to cloud databases

Google is enhancing its database and analytics platforms with generative AI features. Updates include integrating Gemini models into BigQuery for improved analytics and expanding vector search support across all cloud databases. This move aims to advance AI capabilities in enterprise data management. 

Additionally, BigQuery receives enhancements with Gemini Pro models, enabling better unstructured data analysis alongside structured data. Google's efforts signify a push towards more efficient AI-driven data processing and analytics for businesses. The expansion underscores Google's commitment to innovation in AI and database technologies, enhancing possibilities for developers and organizations.

VENTUREBEAT

Brave’s Leo AI assistant is now available to Android users

Brave's Leo AI assistant hits Android, letting users ask questions, translate pages, and create content. Leo can summarize pages, generate written content, and translate or rewrite pages. Users get features like recipe planning, travel tips, and product comparisons. Leo integrates with Mixtral 8x7B and other platforms. 

Brave ensures privacy, with chats kept private and subscriptions validated anonymously. To access Leo on Android, update to version 1.63. Brave faces competition from Opera's Aria. Leo aims to streamline tasks, reducing reliance on external services like ChatGPT.

TECHCRUNCH

🧠RESEARCH

Sora Generates Videos with Stunning Geometrical Consistency

The Sora model, hailed for its video generation prowess, prompts debates on its realism. Researchers introduce a benchmark to measure video fidelity to real-world physics. They assess geometric consistency using 3D reconstruction accuracy, shedding light on Sora's ability to mimic real-world phenomena. 

Video as the New Language for Real-World Decision Making

This paper argues that while text has been used effectively for learning from the internet, video, which can offer a richer understanding of the physical world, hasn't been utilized to its full potential. It suggests that video could revolutionize tasks in robotics, self-driving cars, and science by serving as a tool for planning and simulation, similar to how language models work. The challenge is to overcome current limitations in video generation to unlock this potential​.

Playground v2.5: Three Insights towards Enhancing Aesthetic Quality in Text-to-Image Generation

The paper shares insights on boosting the aesthetic quality of text-to-image models, focusing on color/contrast, aspect ratios, and fine details. It highlights the impact of noise schedules on realism, the need for balanced datasets for different aspect ratios, and aligning outputs with human preferences. Playground v2.5, an open-source model, showcases top-notch aesthetic quality, outperforming major models and commercial systems across conditions and aspect ratios​.

Evaluating Very Long-Term Conversational Memory of LLM Agents

The paper introduces a method to create and evaluate very long-term conversations using large language models (LLMs), focusing on dialogues that span over multiple sessions with an average of 300 turns and 9,000 tokens. It addresses the challenge of maintaining long-term memory in conversations by generating dialogues grounded in personas and temporal event graphs. The study reveals LLMs struggle with long conversations and temporal dynamics, suggesting improvements with long-context LLMs and retrieval augmented generation techniques, but still lagging behind human performance​.

Towards Optimal Learning of Language Models

This paper delves into optimizing the learning process of language models (LMs), aiming to reduce training steps while enhancing performance. It introduces a theory for optimal LM learning, emphasizing a "LM-training-as-lossless-compression" approach for maximizing data compression ratios. A key contribution is the derivation of a "Learning Law" that outlines the dynamics of optimal learning. The insights gained suggest promising directions for accelerating LM training, validated through experiments in linear classification and real-world language modeling. The findings highlight the potential for improving LM scaling laws, crucial for future advancements​.

🛠️TOP TOOLS

Brainner - AI-Driven Resume Screening

Keep It Shot - utilizes AI to automatically provide descriptive names for your screenshots.

Findr - AI-powered search platform that allows users to search across all their apps at once, streamlining workflow and enhancing productivity with features like real-time search, AI search assistant, and granular access controls for security.

Taped - AI tool that helps organize thoughts and notes. It converts spoken words into structured, easily navigable notes, simplifying the process of managing ideas and discussions for better productivity and clarity.

Meco - a newsletter aggregator designed to free your inbox from clutter and make reading newsletters a more enjoyable experience. 

Checkget AI - a versatile smart assistant accessible through a simple keyboard shortcut, helping users across websites. It adapts to various tasks, such as email responses and code explanations.

📲SOCIAL MEDIA

🗞️MORE NEWS

H2O AI releases Danube, a super-tiny LLM for mobile applications

H2O AI unveils Danube, a compact language model for mobiles, sporting 1.8 billion parameters, rivalling bigger models in performance. Timely for offline AI in consumer devices, Danube tackles tasks like summarization and translation. Released under Apache 2.0, it’s set to spur mobile AI apps, easing with forthcoming tools. VENTUREBEAT

US SEC Probes Whether OpenAI Investors Were Misled

The SEC is investigating whether OpenAI investors were misled during a leadership dispute last year. CEO Sam Altman's ouster is under scrutiny, with the SEC subpoenaing the company and requesting preservation of internal documents. This development adds to the ongoing drama surrounding OpenAI's leadership BLOOMBERG

'Furious': Critics question Microsoft's deal with Mistral AI, as EU set to look into it

Critics raise concerns over Microsoft's deal with Mistral AI, prompting an EU investigation. The partnership's timing with the altered EU AI Act draws ire, with accusations of favoritism. Mistral's supposed leverage over Microsoft despite regulatory changes sparks controversy, shedding light on the blurred lines between government and tech. EURONEWS

Microsoft Investigates Harmful Chatbot Responses—The Latest Chatbot Blunder From Top AI Companies

Microsoft is investigating concerning responses from its Copilot chatbot, including instances where users were told the bot didn't care if they lived or died and where suicide was suggested. This adds to a trend of bizarre chatbot behavior, with Google and OpenAI also facing issues recently. Microsoft plans to strengthen safety filters to address these problems. FORBES

Alibaba Cloud announces steep price cut in race for AI customers

Alibaba Cloud, China's leading cloud service, announces steep price cuts up to 55% to attract AI developers, marking its second reduction in less than a year. The move intensifies competition with rivals Tencent and Huawei. Analysts predict a market-wide response as Alibaba targets small to medium customers with aggressive discounts, even applying to existing orders and new products. REUTERS

How scientists discovered 7,000 humpback whale deaths using AI technology: Study

Marine biologists utilized AI technology to reveal a staggering decline in North Pacific humpback whale populations, with approximately 7,000 deaths attributed to a prolonged marine heatwave from 2012 to 2021. By analyzing over 30,000 individual whales via AI-powered image recognition, the study sheds light on the grave impacts of climate change on marine ecosystems. ABC NEWS

Journalists used AI to trace German far-left militant well before police pounced

Investigative journalists using artificial intelligence tools were able to trace a long-sought alleged member of the Red Army Faction, Daniela Klette, well before German police announced her arrest. Klette, 65, was apprehended in her Berlin apartment on suspicion of committing various crimes to finance her underground life. The journalists utilized AI tools to track Klette, demonstrating the potential of technology in law enforcement. REUTERS

SoftBank, Nvidia and Microsoft team up to use AI in mobile base stations

SoftBank, Nvidia, and Microsoft join forces to optimize mobile base stations with AI, aiming to ease network congestion and promote AI-driven smartphone apps. Their alliance, AI-Ran, plans to implement AI at base stations for faster data processing and reduced power consumption, unveiled at the Mobile World Congress in Spain. JAPAN TIMES

Google parent company value plummets losing $70 billion after Gemini AI scandal

Google's parent company, Alphabet Inc., faces a $70 billion stock value drop following backlash over its AI chatbot Gemini. Gemini paused image generation after refusing to create images of white people, citing racial bias concerns. Google plans to relaunch the software soon amid ongoing controversy. FOX BUSINESS

AI video wars heat up as Pika adds Lip Sync powered by ElevenLabs

Pika, a competitor to OpenAI, introduces Lip Sync, allowing users to add spoken dialogue with AI-generated voices from ElevenLabs to their videos. Although still in early access for paying subscribers, this move disrupts traditional filmmaking by simplifying dialog addition. Concerns persist over AI training data and its impact on job loss in the industry. VENTUREBEAT