Sponsored by

Today:

Transform your AI knowledge with The Rundown - the fastest growing AI newsletter in the world

Receive daily insights on cutting-edge AI tools and workflows to boost your productivity, along with the latest and greatest news in the world of AI.

• Trusted by 400,000+ daily readers for the latest in AI, including top industry experts

• Discover AI tools and tutorials that skyrocket efficiency and innovation.

• Stay on top of AI trends and techniques to enhance your work and productivity

OpenAI's Statement SHOCK the Entire Industry! AI Riots vs "Moore's Law for Everything" by Sam Altman

OpenAI's candid warnings, once overlooked or dismissed as distant future concerns, are now acknowledged for their prescience. As AGI's reality dawns, the discourse shifts towards how to navigate the post-AGI world, highlighting the urgent need for adaptation and safe management of AI technologies.

Sam Altman of OpenAI emerges as a central figure, advocating for a future where technological advances in AI could significantly lower living costs and reshape labor dynamics, potentially leading to a more prosperous society for all. This vision hinges on adopting new economic policies that prioritize wealth distribution and inclusivity, proposing a radical overhaul of the current system to harness AI's full potential for humanity's benefit.

WATCH THE VIDEO ON YOUTUBE

Google launches two new open LLMs

Google just dropped Gemma, their new AI brainchild, a week after updating their Gemini lineup. Gemma comes in two sizes, 2B and 7B, ready for commercial and research gigs but keeps the nitty-gritty under wraps, promising only top-tier performance. Unlike fully open-source stuff, Gemma's "open model" vibe means wide access with a catch on customization rights. 

Google's aiming for easy access and use, hooking up with platforms like Hugging Face for smooth integration. They're betting big that Gemma's size and smarts will revolutionize AI app development, pushing the envelope on what's possible with smaller, smarter models. Plus, they're tossing in a toolkit for building AI apps responsibly.

TECHCRUNCH

Enterprise-grade Gemini is now available for teams of all sizes with Google Workspace

Google Workspace has kicked off its Gemini era, transforming Duet AI into Gemini for Workspace. This upgrade integrates Google's top-tier AI into everyday apps, offering a new chat feature with Gemini, backed by solid data protection. There are two subscription options: Gemini Enterprise, starting at $30/user per month for heavy AI users, and Gemini Business at $20/user per month, making advanced AI accessible to all. 

Gemini promises to enhance productivity by aiding in tasks like data analysis and content creation, ensuring privacy and not using data for ads or AI training. This move aims to empower small businesses and large enterprises alike, with plans to extend these benefits to educational institutions soon.

GOOGLE

Nokia Partners With Nvidia on AI in Mobile Networks

Nokia is teaming up with Nvidia to spice up mobile networks with artificial intelligence (AI). They're planning to mix Nvidia's smart chips with Nokia's tech to create better cloud network tech. 

They're also looking into how AI can really change the game for the mobile network industry. This partnership could lead to some cool advancements, making things faster and smarter for everyone using mobile networks.

THE WALL STREET JOURNAL

Intel Launches World’s First Systems Foundry Designed for the AI Era

Intel's making big moves with its Foundry Direct Connect, launching the world's first systems foundry aimed at AI. They're beefing up their tech with the new Intel 14A and other specialized updates, keeping on track to introduce a game-changer: the industry's first backside power solution. 

Big names like the U.S. Commerce Secretary and CEOs from Arm and OpenAI showed up, signaling serious clout. They've got Microsoft on board, choosing Intel's 18A process for a new chip, plus a bunch of ecosystem partners like Synopsys and Cadence are ready to roll with tools and IP to support Intel's push. 

Intel's laying out a roadmap that could see them leading the pack by 2025, with new nodes and packaging tech coming down the line. They're not just about chips; they're aiming to be the most sustainable foundry around, using mostly renewable energy and targeting zero waste. With over $15 billion in the pipeline, Intel's betting big on becoming a key player in AI and beyond, promising to mix it up in tech and sustainability.

INTEL

AI Model Backed by Asia’s Richest Person to Launch in March

Mukesh Ambani, Asia's wealthiest person, is throwing his weight behind BharatGPT's Hanooman service, set to debut next month. This launch is a major move for India, showcasing its drive to be a key player in the AI scene. 

A team-up between Reliance Industries and India's elite engineering colleges is making it happen, signaling big ambitions and a competitive edge in the global AI race. 

BLOOMBERG

Match Group inks deal with OpenAI, says press release written by ChatGPT

Match Group, known for dating apps like Tinder and OkCupid, announced a partnership with OpenAI, securing over 1,000 enterprise licenses. This $20 million AI investment in 2024 aims to boost employee productivity by assisting in tasks such as coding and customer communications. 

Despite the cheesy press release, courtesy of ChatGPT, highlighting Match's commitment to AI safety and productivity, the initiative marks a significant bet on AI to enhance app functionality and user safety. Match plans to roll out AI across its platforms, promising improved matchmaking and user experiences, while exploring new AI-driven apps. The move underscores AI's growing influence in enhancing business operations and user engagement in the digital dating landscape.

TECHCRUNCH

ChatGPT starts spouting nonsense in 'unexpected responses' shocker

OpenAI's ChatGPT started spewing nonsense last night, leaving users scratching their heads and sharing their bizarre encounters online. From poetic ramblings about computers to speaking in tongues, ChatGPT's "word salad" had folks wondering what went wrong. Turns out, a setting tweak behind the scenes might have thrown ChatGPT off its game. 

OpenAI quickly jumped on the issue, fixing it within minutes, and now ChatGPT is back to giving coherent answers. Despite the hiccup, this incident has sparked a conversation about the reliability of AI, with experts calling for more transparent, manageable, and understandable AI systems. OpenAI is keeping an eye on things, but hasn't spilled the beans on what exactly caused the glitch.

THE REGISTER

🧠RESEARCH

Neural Network Diffusion

The paper discusses a groundbreaking method where diffusion models, initially celebrated for their prowess in creating images and videos, are now used to generate high-quality neural network parameters. By combining an autoencoder with a standard latent diffusion model, this technique synthesizes new network parameters from noise, creating models that either match or outdo traditionally trained networks in performance without extra hassle. This innovation opens new avenues for utilizing diffusion models beyond their conventional applications.

Synthetic Data (Almost) from Scratch: Generalized Instruction Tuning for Language Models

GLAN is a game-changer in language model training, which generates a vast pool of synthetic instruction data from scratch. It reimagines instruction tuning by crafting a detailed taxonomy of human knowledge, without relying on specific examples or existing datasets. By breaking down knowledge into fields and disciplines, GLAN covers the whole spectrum of human intellect to train language models. This approach shines in various tasks like math, coding, and logical reasoning, showcasing its versatility and the potential for easy updates with new knowledge areas.

VideoPrism: A Foundational Visual Encoder for Video Understanding

VideoPrism is a groundbreaking video encoder designed for a wide range of video understanding tasks using a single, unchanging model. VideoPrism is pre trained on a mix of 36 million high-quality video-caption pairs and 582 million video clips with text, like ASR transcripts. This training method, which includes masked autoencoding with global-local distillation and token shuffling, allows VideoPrism to focus on video content while utilizing associated text. It sets new benchmarks, outperforming previous methods in 30 out of 33 video understanding tasks.

Improving Robustness for Joint Optimization of Camera Poses and Decomposed Low-Rank Tensorial Radiance Fields

This paper presents an innovative algorithm that refines camera poses and scene geometries using 2D images. By examining 1D signals and extending findings to 3D, the study highlights the pitfalls of naive optimization in voxel-based Neural Radiance Fields (NeRFs) and introduces a convolutional Gaussian filter-based method for efficient, robust optimization. This approach, enhanced with techniques like smoothed supervision and edge-guided loss, significantly improves novel view synthesis and optimization speed.

Video ReCap: Recursive Captioning of Hour-Long Videos

"Video ReCap" revolutionizes video captioning by handling hour-long videos and generating captions across multiple hierarchy levels. It employs a recursive architecture to process videos of varying lengths efficiently, from brief clips to 2-hour features. Through a curriculum learning approach, it gradually understands video hierarchies, starting with clip-level actions to segment-level descriptions, culminating in summaries for extensive footage. Additionally, it introduces the Ego4D-HCap dataset, enhancing Ego4D with long-range video summaries. This model not only flexibly captions at various levels but also aids complex video understanding tasks.

🛠️TOP TOOLS

Imprompt - offers a suite of business-grade chat and generative AI management solutions, focusing on privacy and efficiency. 

Geppetto - advanced Slack bot that integrates OpenAI's ChatGPT-4 and DALL-E-3 for interactive AI conversations and image generation. 

MetaVoice - text-to-speech technology, offering a platform to experiment with various voice styles and settings. 

Sintra AI - automating tasks and processes, aiming to fuel business growth through the use of AI prompts and automation bots. 

Adventure AI - educational platform designed to teach kids real AI skills through a social game.

📲SOCIAL MEDIA

🗞️MORE NEWS

Shining Brighter Together: Google’s Gemma Optimized to Run on NVIDIA GPUs

NVIDIA and Google joined forces to optimize Gemma, Google's latest open language models, for NVIDIA AI platforms. Gemma, with 2B and 7B parameters, is tailored for various uses, running efficiently on over 100 million NVIDIA RTX GPUs worldwide. Developers can leverage NVIDIA's tools for fine-tuning and deploying Gemma in production applications. NVIDIA

'AI godfather', others urge more deep fake regulation in open letter

Artificial intelligence experts, including Yoshua Bengio, advocate for stricter regulation of deepfakes due to potential societal risks. Deepfakes, realistic yet fabricated content created by AI, pose threats such as fraud and disinformation. The letter suggests criminalizing harmful deepfakes and enforcing accountability for AI companies. Over 400 individuals from various sectors have endorsed the call for regulation. REUTERS

Chinese fintech firm claims priority access to Sora API through Microsoft amid new text-to-video AI frenzy

A Chinese fintech firm, Sinodata, is leveraging its partnership with Microsoft to secure priority access to OpenAI's Sora API, a text-to-video tool. This move underscores the rush among Chinese businesses to align with cutting-edge AI technologies like Sora, despite challenges in direct access within the country's regulatory framework. SCMP

AI influencers are making their secretive creators tens of thousands of dollars a month—now an OnlyFans rival is betting on the lucrative virtual girlfriends

Tech-savvy creators are cashing in on AI influencers, resembling OnlyFans models. These faceless personas, controlled by creators, engage paying subscribers on social media. Fanvue, an OnlyFans competitor, is banking on this trend, with AI models doubling revenue monthly. Despite secrecy, AI-generated personalities like Emily Pellegrini rake in thousands monthly. FORTUNE

Scientists say they can use AI to solve a key problem in the quest for near-limitless clean energy

Scientists have made progress in nuclear fusion energy research by using artificial intelligence (AI) to predict and prevent plasma instabilities. Fusion energy, seen as a clean energy solution, has long been hindered by technical challenges. AI offers a promising solution to overcome these obstacles, bringing fusion energy closer to reality. CNN

How a Lisbon startup's AI forecasting tech outperforms Google, Amazon, and Meta

Lisbon startup AssetFloow's AI forecasting tech surpasses giants Google, Amazon, and Meta, offering 20% better accuracy using 50% less data. Their Dynamic Flow Network aids retailers with precise shopper behavior insights, reducing waste and boosting sales. Co-founder Ricardo Santos unveils DFN's versatility and success, aiming to tackle retail's $1.5 trillion inventory issue. TECH.EU

Nearly all doctors feel burnout, but many are encouraged by potential of AI, survey says

US doctors are overwhelmed by burnout, paperwork, and staff shortages, with over 90% feeling burnt out regularly. Despite this, 83% believe AI could help alleviate these issues. Administrative burdens are cited as the main cause of burnout, prompting physicians to work extra hours. Many see potential for AI to streamline tasks and improve patient care, though concerns about losing the human touch remain.  CNBC

5 AI Accounting Tools For Freelancers Running A Side Hustle In 2024

Freelancers can alleviate accounting stress with AI tools. Dext automates expense tracking, especially for e-commerce. Booke AI learns categorization habits, integrates with accounting software, and provides reconciliation advice. TurboTax offers expert assistance and filing support. QuickBooks Self-Employed estimated quarterly taxes, while FreshBooks includes time tracking for productivity. FORBES

💰FUNDING

BRIA raises $24 million Series A for ‘ethical’ visual generative AI platform

India’s Tech Leaders Jockey For Slice of $300 Billion AI Market

Loora lands $12 million Series A for generative AI English learning platform

Bioptimus raises $35 million seed round to develop AI foundational model focused on biology