Humanoid AI revolution

PLUS: Apple Challenges Microsoft's AI, Groq Unveils GroqCloud Platform and more.


Microsoft's new "Embodied AI" SHOCKS the Entire Industry! | Microsoft's Robots, Gaussian Splat & EMO

In the bustling world of AI, a collaboration between Figure AI, Microsoft, and OpenAI has set the tech community abuzz. With hefty investments from giants like Nvidia and Jeff Bezos, Figure AI is on a fast track to revolutionizing humanoid robots. Their partnership with OpenAI aims to push the boundaries of AI models, while Microsoft's Azure lends its cloud muscle for AI training and storage. 

This trio's synergy promises to speed up the deployment of humanoid robots, signaling a leap towards commercial AI applications. A recent demo showcased a robot's impressive autonomy, far surpassing previous attempts, and hinting at a future where AI seamlessly integrates into our daily lives. This venture not only underscores the rapid advancements in AI but also highlights the strategic moves by leading tech companies to shape the future of robotics and AI infrastructure.

Elon Musk Sues OpenAI and Altman for Breaching Firm’s Founding Mission

Elon Musk sues OpenAI, alleging betrayal of their founding agreement to develop AI for humanity's benefit, not profit. Musk claims OpenAI's collaboration with Microsoft has turned it into a closed-source subsidiary focused on profit, rather than open-source advancement. He accuses OpenAI of breaching contract and fiduciary duty, aiming to revert it to open source and prevent profit from its technology. 

Musk highlights the closed nature of GPT-4, OpenAI's latest AI model, and criticizes the board's lack of technical expertise. He fears this shift may compromise public safety and accuses Microsoft of influencing OpenAI to prioritize profit over openness.

Apple's Launching A Futuristic Siri And Apple-Based Artificial Intelligence As It Looks To Take On Microsoft's AI Dominance

Apple Inc. is shifting focus from its long-awaited Apple Car project to artificial intelligence (AI). With a hefty $73 billion at its disposal, Apple plans to unveil generative AI tools at its Worldwide Developers Conference. These innovations include an upgraded Siri and an AI-based code completion tool for developers, aiming to enhance user experience and productivity. 

Apple's move mirrors Microsoft's AI dominance, with both giants investing heavily in AI technology. The shift signifies Apple's ambition to integrate AI across its product range, signaling a new era in consumer technology.

Stack Overflow to charge LLM developers for access to its coding content

Stack Overflow has introduced an API requiring AI models to credit its content and pay for access. Google partners with Stack Overflow to enhance Gemini models. Overflow API aids code generation and answers queries, integrated with Google Cloud Console. Stack Overflow asserts compensation for its data. 

Amid AI-driven solutions, Stack Overflow emphasizes human-generated content's relevance for fresh data and community context. Despite declining traffic, Stack Overflow sees its role as pivotal for AI development. With a focus on community attribution and data quality, Stack Overflow aims to maintain its significance in the AI era.

Groq launches developer playground GroqCloud with newly acquired Definitive Intelligence

Groq, a California-based startup known for its speedy microchips for AI, made headlines by acquiring Definitive Intelligence. Together, they launched GroqCloud, a platform for developers with easy access to tools and docs. Their main draw? The Groq Language Processing Unit (LPU) Inference Engine, praised for its lightning-fast results. 

While Nvidia's GPUs dominate, Groq aims to lead with its purpose-built LPUs. GroqCloud soft-launched recently, attracting thousands of developers. The move aligns with their vision for an accessible AI economy. This acquisition also birthed Groq Systems, targeting the public sector. Groq is carving a path for AI accessibility and innovation, challenging industry giants like Nvidia.


The BigCode project teamed up with Software Heritage to launch StarCoder2, a beefed-up code-crunching AI that's trained on a massive pile of code from 619 programming languages. This new version is four times larger than its predecessor and comes in different sizes, with the smallest model punching above its weight against similar-sized rivals and even its own bigger sibling. The largest model not only outshines others in its weight class but also holds its own against giants twice its size, especially in math, code reasoning, and languages not widely used. 

Griffin is a cutting-edge approach that blends gated linear recurrences with local attention to create efficient language models. It outperforms existing models in speed and can handle longer sequences than seen during training, offering high throughput and low latency. Despite using significantly fewer training tokens, Griffin rivals top models in performance, demonstrating a breakthrough in model efficiency and scalability.

Panda-70M is a game-changer in video captioning, by automatically generating captions for 70 million videos using multimodal inputs and cross-modality teacher models. This massive dataset leapfrogs manual labeling hurdles, offering a treasure trove for improving video captioning, retrieval, and text-driven video generation tasks. The novel method not only scales up video-text pair creation but also boosts performance across key benchmarks, showcasing the dataset's immense potential for advancing multimedia research.

bGPT is a pioneering model that goes beyond traditional deep learning by focusing on bytes, the fundamental units of the digital world. By predicting the next byte, bGPT excels in various modalities like text, audio, and images, offering a new way to predict, simulate, and diagnose digital processes. It achieves remarkable accuracy in music data conversion and CPU behavior simulation, showcasing its ability to learn directly from binary data and simulate complex digital patterns.

The paper presents a novel approach to humanoid locomotion, treating it as a next token prediction problem similar to word prediction in language. By using a causal transformer to predict sensorimotor trajectories, the model is trained on a mix of simulated trajectories, motion capture data, and real-world videos. It demonstrates that a full-sized humanoid can walk in San Francisco without prior specific training, showing promise for applying this technique to real-world control tasks through generative modeling of sensorimotor pathways.


NLUX - The Open-Source Javascript Library To Build Conversational AI Interfaces

Fabric - open-source framework for augmenting humans using AI.

Deepgram - a conversational text-to-speech API for voice AI agents. It aims to deliver human-like conversation quality at high speed and efficiency. 

Conveyor - GPT-powered automation software simplifies security questionnaires with over 90% accuracy, saving infosec teams significant time.

Fathom - free AI Meeting Assistant that records, transcribes, and summarizes meetings, supporting multiple languages.

Cape Privacy - a privacy-focused Conversational AI platform with various AI models, including GPT-3.5 Turbo and GPT-4.



Hugging Face, the GitHub of AI, hosted code that backdoored user devices

Hugging Face, like GitHub for AI, unknowingly hosted AI code containing hidden malware, exposing users to backdoors and other threats. JFrog researchers discovered around 100 malicious submissions, including one that granted remote control of devices. Using Python's risky "pickle" format, attackers injected code into models, highlighting the need for careful vetting on AI platforms. ARSTECHNICA

Vimeo launches new AI video tools to help employees breeze through hours-long town halls and training videos. Will it usher in the golden age of asynchronous work?

Vimeo launches new AI video tools to improve corporate training and town halls, aiming to ease the burden on employees. By utilizing generative AI, Vimeo seeks to enhance the experience of lengthy videos, potentially revolutionizing asynchronous work methods. FORTUNE

US military operators started out skeptical about AI, but now they are the ones developing and using Project Maven to identify targets on the battlefield.

In 2020, the U.S. Army's 18th Airborne Corps made history at Fort Liberty by using AI to pinpoint and destroy a target, marking the U.S.'s first AI-enabled strike. Fast forward, Project Maven, the Pentagon's AI initiative, now plays a crucial role in identifying military targets worldwide, including in Yemen, the Red Sea, Iraq, and Syria. Despite initial skepticism and technical challenges, the technology has advanced, with AI now integral to the U.S. military strategy, speeding up decision-making and targeting processes. BLOOMBERG

Mark Zuckerberg's Asia trip is all about making AI friends

Mark Zuckerberg's Asia tour isn't just about sightseeing and fast food. Amidst skiing and sushi, he's hustling hard, meeting political and business bigwigs to bolster Meta's AI and mixed reality ventures. With competitors looming and regulatory challenges ahead, Zuckerberg's on a mission to secure Meta's future in the region. BUSINESS INSIDER

Microsoft showcases next-generation AI and computing at Nvidia GTC

Microsoft highlights practical AI applications and innovations at Nvidia GTC, emphasizing real ROI and diverse business opportunities. They focus on enriching employee experiences, customer engagement, business process enhancement, and innovation acceleration. Through sessions and demos, they showcase Azure AI's capabilities and offer hands-on training, aiming to empower businesses of all sizes in AI adoption. VENTUREBEAT

OpenAI rival Cohere says some AI startups build Bugatti sports cars, ‘We make F-150s’

Cohere, an AI startup backed by Nvidia and founded by ex-Google AI researchers, focuses on enterprise generative AI rather than consumer chatbots. Dubbed the "F-150s" of AI, Cohere aims to provide practical solutions for businesses, avoiding the glamour of consumer-focused AI akin to "fancy sports cars." With a recent $270 million funding round and partnerships with industry giants like Salesforce and Oracle, Cohere aims to expand its enterprise-only chatbots while maintaining independence from conditional investments. Despite the surge in generative AI interest, Cohere stands out with its enterprise-focused approach, targeting areas like search and retrieval for deployment at scale in 2024. CNBC

"Forty-one per cent of architects now using AI" says RIBA report

A recent report by RIBA finds that 41% of UK architects are using artificial intelligence in their projects, with 43% seeing improved efficiency. Despite concerns about job security and creativity, RIBA President Muyiwa Oki believes AI is shaping the future of architecture irreversibly, urging responsible innovation. DEZEEN

Google's new math app solves nearly any problem with AI: Here's how to use it

Google's Photomath, now part of the Google app family, utilizes AI to swiftly solve math problems from basic arithmetic to advanced calculus. With over 100 million downloads, its accuracy and speed make it a top choice for students and parents alike, offering step-by-step explanations for learning. ZDNET

What'd you think of today's edition?

Login or Subscribe to participate in polls.

What are MOST interested in learning about AI?

What stories or resources will be most interesting for you to hear about?

Login or Subscribe to participate in polls.

Join the conversation

or to participate.