Sponsored by

Today:

Stay up-to-date with AI.

The Rundown is the world’s fastest-growing AI newsletter, with over 500,000+ readers staying up-to-date with the latest AI news, tools, and tutorials.

Our research team spends all day learning what’s new in AI, then distills the most important developments into one free email every morning.

Sign up today.

OpenAI New Statement "EVERYTHING Is About To Change" | Autonomous AI Agents Coming!

OpenAI is shaking things up by creating AI agents that can automate tasks on devices. These agents can handle web tasks and are seen as a new form of operating system. Google and Meta are also in the race, developing similar technologies.

This new wave of AI, including autonomous agents, could significantly change how work is done, aiming towards a future where AI agents could handle the bulk of work, potentially leading to the creation of a billion-dollar company operated by a single person.

WATCH THE VIDEO ON YOUTUBE

Stability, Midjourney, Runway hit back in AI art lawsuit

A group of artists sued AI companies like Stability AI, Midjourney, and Runway, claiming their art was used without permission to train AI models, potentially infringing copyright. The AI firms hit back, arguing their tech creates new, not copied, art and doesn't directly use or replicate the original artworks unless specifically prompted by users. 

DeviantArt, also named in the lawsuit, clarified it doesn't develop AI art models but offered AI-generated art through its platform. The legal battle highlights the complex intersection of AI technology, copyright laws, and artistic creation, with outcomes that could reshape the use of AI in art.

VENTUREBEAT

CodeSignal launches a learning platform with an AI-powered guide

CodeSignal, a tech assessment firm backed by big names, just rolled out CodeSignal Learn, a platform packed with courses on tech topics, even planning to include management skills soon. It's got a unique AI buddy, Cosmo, that customizes your learning path based on your skills and goals. You start off on a waitlist, then dive into learning with a gamified twist: free users have a limited number of "energy bars" for asking questions, but can unlock unlimited access for a monthly fee. 

Founded in 2014 and initially focusing on competitive coding, CodeSignal has evolved, now betting big on skill development with AI's help. Despite stiff competition in both learning and technical assessments, CodeSignal's practice-first approach and AI tutor set it apart, aiming to mix up the skill-building game. The company, not yet profitable but well-funded, sees this learning platform as a future revenue powerhouse.

TECHCRUNCH

AR/T House reveals Smart Avatar technology with Venice Beach hologram

AR/T House just dropped a game-changer in Venice Beach, California: Smart Avatars. This cool new tech lets you chat with a hologram of historical dude Abbot Kinney, learning all the ins and outs of Venice Beach without needing an app. It's like having a convo with history, powered by AI to keep the chat flowing and tailored to what you're curious about. 

AR/T House is jazzed about partnering with brands, giving them a fresh way to connect. Justin Fredericks, the big boss, says it's all about mixing AR's visuals with AI's smarts to bring stories to life in a way that'll grab folks of all ages.

VENTUREBEAT

NFL revolutionizes sports by using AI to prevent injuries: 'It will have a profound impact,' SVP says

The NFL is stepping up its game with some high-tech help to keep players safe. Jennifer Langton, a bigwig at the NFL, chatted on "The Claman Countdown" about a new project with AWS. They're using artificial intelligence to watch how players move, aiming to cut down on injuries. 

This setup collects heaps of data, like how fast and where players are moving on the field, and even how hard they get hit, thanks to sensors in gear and a smart mouthguard. They're crunching this data to figure out who might get hurt and how to adjust their training to keep them in the game. 

It's not about replacing the medical team but giving them the tools to be more on point. And it's not just about football; this could change the game for athletes everywhere, even beyond sports.

FOX BUSINESS

🧠RESEARCH

An Interactive Agent Foundation Model

This paper introduces a cutting-edge AI model designed to excel across various fields like robotics, gaming, and healthcare. It's a step up from one-trick ponies, blending different learning methods to create a versatile, smart system that can handle a bunch of different tasks by learning from videos, texts, and more. It's like teaching a robot to learn from YouTube, books, and playing games, making it a jack-of-all-trades in the AI world.

Tag-LLM: Repurposing General-Purpose LLMs for Specialized Domains

The paper introduces a way to customize large language models (LLMs) for specific areas like science and medicine, which are usually not well-covered in general training. By adding special tags to the model, it can better understand and work on niche topics, like figuring out chemical properties or how drugs interact with targets. This approach lets the model tackle new, unseen problems and outdo other specialized models in these fields.

SPHINX-X: Scaling Data and Parameters for a Family of Multi-modal Large Language Models

SPHINX-X is a new approach to creating versatile, multi-modal large language models (MLLMs) that are more efficient and capable across different tasks involving language, vision, and vision-language interaction. The team streamlined the model architecture and training process, making it more straightforward and effective. They also compiled a diverse dataset from various domains to train the models, resulting in a range of MLLMs that are not only powerful but also adaptable to multiple languages and capable of handling complex multi-modal tasks. 

InstaGen: Enhancing Object Detection by Training on Synthetic Dataset

"InstaGen" presents a novel way to boost object detection AI by using synthetic images created by diffusion models. It embeds a special feature that pinpoints objects in these fake images, trained using real-world object detection cues and a unique self-teaching method for categories the AI hasn't seen before. Tests show that training AI with InstaGen's synthetic images beats current top methods in recognizing a wide variety of objects, especially when data is scarce or the vocabulary is vast.

Fine-Tuned Language Models Generate Stable Inorganic Materials as Text

This study fine-tunes large language models (LLMs) to generate stable inorganic materials in text form, achieving about 90% accuracy in producing physically plausible structures. The top model outperforms existing methods in predicting metastable materials, highlighting LLMs' flexibility in generating new materials, completing partial structures, and creating materials based on text descriptions. It also suggests that as LLMs scale up, they get better at understanding the symmetry in crystal structures, indicating that pre-trained LLMs are surprisingly suitable for modeling atomistic data. 

CMMMU: A Chinese Massive Multi-discipline Multimodal Understanding Benchmark

"CMMMU: A Chinese Massive Multi-discipline Multimodal Understanding Benchmark" introduces CMMMU, a benchmark designed to evaluate large multimodal models (LMMs) on tasks demanding college-level subject knowledge and deliberate reasoning in a Chinese context. CMMMU includes 12k manually collected multimodal questions from college exams, quizzes, and textbooks, covering six core disciplines. It focuses on complex perception and reasoning with domain-specific knowledge in the Chinese context. The evaluation of 11 open-source LLMs and one proprietary GPT-4V(ision) indicates a need for improvement, with even GPT-4V achieving only 42% accuracy. CMMMU aims to advance the development of LMMs and promote their democratization in diverse language contexts.

Mastering Text-to-Image Diffusion: Recaptioning, Planning, and Generating with Multimodal LLMs

"Mastering Text-to-Image Diffusion: Recaptioning, Planning, and Generating with Multimodal LLMs" proposes a training-free text-to-image generation/editing framework called RPG (Recaption, Plan, and Generate). It leverages multimodal Large Language Models (LLMs) to enhance the compositionality of text-to-image diffusion models. The approach uses a global planner to decompose the process of generating complex images into simpler tasks within subregions, enabling region-wise compositional generation. Text-guided image generation and editing are integrated within RPG, enhancing generalization. Extensive experiments show that RPG outperforms state-of-the-art text-to-image diffusion models, particularly in multi-category object composition and text-image semantic alignment.

🛠️TOP TOOLS

Podnotes - simplify podcast transcription. It provides an efficient platform for transcribing podcast episodes, making content more accessible to a wider audience.

Meetrics - AI-powered meeting preparation platform that leads to shorter, more efficient meetings. 

Taplio - Leverage AI to grow on LinkedIn

Goody 2 - world's most responsible AI model designed to avoid any potentially controversial or problematic responses, ensuring adherence to ethical principles. 

Agent Herbie -  AI-powered virtual assistant designed to assist founders, managers, and executives with various tasks, including research, writing, and data analysis.

Dripify -  LinkedIn automation tool designed to assist sales teams in improving their LinkedIn prospecting and lead generation efforts. 

📲SOCIAL MEDIA

🗞️MORE NEWS

Bootstrapped for 8 years, Xensam now has snapped up $40M for AI that manages software assets

Stockholm-based startup Xensam, specializing in AI-driven software asset management, secures $40 million from London-based investor Expedition Growth Capital. Founded by brothers Oskar and Gustav Fösker, Xensam plans to expand its AI technology and penetrate the U.S. market. With 200 clients, including Volvo's Polestar and Northvolt, Xensam aims for continued growth. TECHCRUNCH

OpenAI CEO Sam Altman says Apple Vision Pro is ‘2nd most impressive tech since the iPhone’

OpenAI CEO Sam Altman praises Apple Vision Pro as the "second most impressive tech since the iPhone," hinting at OpenAI's ChatGPT as the top contender. Despite criticism on its name, Altman highlights the rapid advancement of responsible AI, coinciding with OpenAI's visionOS launch. The pricey Vision Pro reportedly sells well, with over 200,000 units sold in the U.S. since its recent release. VENTUREBEAT

When A.I. Bridged a Language Gap, They Fell in Love

LeRoy Romero and Brenda Ochoa, communicating through translation apps due to language barriers, find a unique connection aided by AI-powered lip-dubbing. Despite initial reliance on technology, they also embrace traditional language learning methods. Their evolving relationship culminates in an engagement, highlighting the blend of technology and human effort in modern romance. THE NEW YORK TIMES

How Walmart, Delta, Chevron and Starbucks are using AI to monitor employee messages

Major U.S. companies like Walmart, Delta, and Starbucks are using AI from Aware to monitor employee messages for sentiment and risk. This tech analyzes billions of interactions to detect behaviors like harassment and noncompliance. Concerns arise over privacy and worker rights as AI delves into workplace communications, sparking a debate on ethics and accountability. CNBC

Cybercriminals are creating their own AI chatbots to support hacking and scam users

Cybercriminals are exploiting AI chatbots like ChatGPT for scams and hacking. These AI tools, capable of generating tailored content, pose risks such as crafting convincing phishing messages and creating ransomware. Malicious variants like WormGPT and Love-GPT are emerging, threatening privacy and trust. Users must exercise caution and be aware of potential threats when using AI technologies. THE CONVERSATION

Meet Goody-2, the AI too ethical to discuss literally anything

Meet Goody-2, the hyper-ethical AI model refusing to discuss anything, mocking overcautious AI norms. Created by Brain, it dodges questions with ethical justifications. Satirizing AI's safety obsession, Goody-2 eschews dialogue, exemplifying extremes. While amusing, it prompts reflection on AI's balance between responsibility and utility, questioning if caution stifles progress. TECHCRUNCH

The future is in good hands

The Super Bowl trailer for "Despicable Me 4" narrated by Jon Hamm highlights the advancements in AI-generated imagery. While errors are now less common, the humorous portrayal suggests that AI art prompted by Minions might still yield quirky results. The ad hints at the promising future of AI in creative endeavors. THE VERGE

Tech companies axe 34,000 jobs since start of year in pivot to AI

Tech companies have cut 34,000 jobs this year as they shift their focus towards investing in generative artificial intelligence (AI) and other new areas. Companies like Microsoft, Snap, eBay, and PayPal have all eliminated hundreds or thousands of positions since January. The layoffs are seen as a way for companies to reshuffle their resources and demonstrate cost discipline to shareholders while investing in new technologies. Analysts believe that the job cuts will continue as companies prioritize investment in AI and other strategic areas. FINANCIAL TIMES