Today:

• Lambda Labs' $800M Goal

• Rain AI Hires Apple Chip Expert

• Hebbia's $100M Series B

• Detect-2B's Impressive Accuracy

• Nokia's $2.3B AI Bet

• ElevenLabs Unveils IOS App

Nvidia partner Lambda Labs seeks $800mn as AI computing demand soars

Lambda Labs, a partner of Nvidia, is aiming to secure $800 million in funding to meet the increasing demand for AI computing solutions. The company offers a range of subscription packages that include access to global news & analysis, expert opinions, FT apps, newsletters, and special features.

Customers can choose from various subscription tiers, such as Standard Digital, Premium Digital, and Weekend Print delivery, each offering different levels of access and benefits. In addition to providing digital editions and exclusive analyses, Lambda Labs also caters to occasional readers with free access and offers volume discounts for businesses.

FINANCIAL TIMES

Sam Altman-Backed Rain AI Hires Apple Chip Vet to Lead Hardware

Rain AI, a startup backed by OpenAI's Sam Altman, has hired Jean-Didier Allegrucci, a former Apple chip expert, to lead its hardware division. This is the company's second major hire this month, following the recruitment of Amin Firoozshahian from Meta. Rain AI aims to create more efficient chips for artificial intelligence using in-memory computing, which processes data where it's stored, similar to how human brains work. 

This approach could make AI systems faster and use less energy. The company is competing with big tech firms like Intel and Samsung in this field. Rain AI's goal is to make AI technology more accessible and affordable for widespread use.

BLOOMBERG

Hebbia raises nearly $100M Series B for AI-powered document search led by Andreessen Horowitz

Hebbia, a startup using artificial intelligence to search large documents and provide answers, has raised nearly $100 million in new funding. The investment was led by Andreessen Horowitz, a well-known venture capital firm. Hebbia's value is now estimated between $700-800 million. 

The company was started in 2020 by George Sivulka to help people in finance save time when reading long documents. Hebbia's technology can analyze many types of files at once and give specific answers. It's mainly used by financial companies but could be useful in other fields too. This new money brings Hebbia's total funding to over $120 million.

TECHCRUNCH

Resemble AI’s next-generation AI audio detection model, Detect-2B, is 94% accurate

Resemble AI has released Detect-2B, a new model for identifying AI-generated audio with 94% accuracy. This tool uses pre-trained sub-models to examine audio clips for signs of artificial creation. It focuses on subtle sounds that often distinguish real from fake recordings. The model's design allows it to adapt to new audio without retraining. Detect-2B uses a flexible approach that works well with various audio qualities and languages. 

As concerns grow about AI-generated voices misleading voters or eroding trust in brands, tools like this become increasingly important. Resemble AI plans to make Detect-2B available through an API for integration into different applications, while continuing to improve its capabilities.

VENTUREBEAT

Nokia taps AI boom with $2.3 bln Infinera purchase

Nokia is buying U.S. company Infinera for $2.3 billion to boost its position in optical networking, which is crucial for data centers and AI growth. This move will make Nokia the second-largest player in the field, behind Huawei. The deal helps Nokia diversify as 5G equipment sales decline, allowing it to supply more gear to tech giants building data centers. 

Nokia's CEO sees this as timely, coinciding with market recovery and AI-driven investments. The purchase is mostly cash, with some stock, and Nokia expects to save 200 million euros after the deal closes. Investors reacted positively, with Nokia's shares rising 4% in early trading.

REUTERS

ElevenLabs launches iOS app that turns ‘any’ text into audio narration with AI

ElevenLabs, an AI audio company, has released an iOS app that turns text into lifelike speech. The app, called "ElevenLabs Reader: AI Audio," allows users to listen to articles, books, or documents using AI-generated voices. It offers features like adjustable reading speed and various voice options. Users can input text by writing, pasting links, or uploading files. 

The app works with different content types and integrates with iOS sharing features. While some initial bugs were reported, the AI narration was praised for its accuracy and natural sound. The app is free to download but requires an ElevenLabs account. Questions remain about data privacy and storage of user-submitted content. 

VENTUREBEAT

🧠RESEARCH

OMG-LLaVA: Bridging Image-level, Object-level, Pixel-level Reasoning and Understanding

OMG-LLaVA is a new framework that combines pixel-level vision understanding with reasoning abilities. It integrates image and text prompts for flexible user interaction. Using a universal segmentation method as the visual encoder, it achieves image, object, and pixel-level reasoning. The model surpasses specialized methods and is available for research.

Step-DPO: Step-wise Preference Optimization for Long-chain Reasoning of LLMs

Step-DPO improves mathematical reasoning in large language models (LLMs) by optimizing individual reasoning steps rather than whole answers. This method uses a dataset of 10,000 step-wise preference pairs, showing that self-generated data enhances performance. Step-DPO increases accuracy on benchmarks, surpassing closed-source models like GPT-4. The code and data are publicly available.

MUMU: Bootstrapping Multimodal Image Generation from Text-to-Image Data

MUMU is a model that generates images from mixed text and image prompts. It uses a vision-language encoder with a diffusion decoder, trained on semantically meaningful image crops. MUMU can blend elements from different images into a cohesive output, enabling tasks like style transfer and character consistency. The model demonstrates the potential of multimodal models for versatile image generation.

CharXiv: Charting Gaps in Realistic Chart Understanding in Multimodal LLMs

CharXiv is a new evaluation suite for assessing chart understanding in Multimodal Large Language Models (MLLMs). It features 2,323 complex, real-world charts from arXiv papers and includes both descriptive and reasoning questions. Results show a significant performance gap between human accuracy (80.5%) and top models like GPT-4o (47.1%) and InternVL Chat V1.5 (29.2%). This highlights the need for improved chart understanding in MLLMs.

DiffusionPDE: Generative PDE-Solving Under Partial Observation

DiffusionPDE is a framework that uses generative diffusion models to solve partial differential equations (PDEs) under incomplete observations. Traditional methods struggle with partial data, but DiffusionPDE models the joint distribution of solutions and coefficients to fill in missing information and solve the PDE. This approach significantly outperforms current methods in both forward and inverse PDE-solving scenarios.

🛠️TOP TOOLS

Eve Coach - AI Stress coach

ElevenLabs Video to Sound Effects - Generate custom AI sound effects for your videos

Magnific AI - Most advanced AI tech to achieve insanely high-res upscaling

Character AI - Generates life-like conversations between you and any character

Figma AI - Get started faster, find what you’re looking for, and stay in the flow

🗞️MORE NEWS

In a Surprise, OpenAI Is Selling More of Its AI Models Than Microsoft Is

OpenAI's AI model sales now surpass Microsoft's. Initially, Microsoft partnered with OpenAI in 2019 to leverage its sales network. Despite lacking a strong sales team until last year, OpenAI has excelled, outpacing Microsoft's AI model revenue, a notable achievement given Microsoft's dominance in enterprise software. THE INFORMATION

Toys 'R' Us released an AI-generated ad. It's getting mixed reviews

Toys "R" Us released an AI-generated ad using OpenAI's Sora tool, depicting its late founder Charles Lazarus as a child. The ad, created with Native Foreign, received mixed reviews, highlighting both impressive AI capabilities and flaws. This move reflects the broader impact of generative AI on creative industries. QUARTZ

Perplexity AI: the answer engine with a lot of question marks

Perplexity AI, an "answer engine" startup, faces accusations of plagiarism and ignoring web crawling rules. Despite funding from Jeff Bezos and aiming to rival Google, the company draws media ire for using AI-generated content and paywalled sources without permission. Perplexity's practices highlight ethical issues in the AI industry. THE VERGE

AI can beat university students, study suggests

A study by the University of Reading found that AI-generated exam answers from ChatGPT outperformed real students, scoring half a grade higher on average. AI-generated essays were mostly undetected by markers, raising concerns about academic integrity. The findings suggest educators need to adapt assessment methods in response to AI's impact. BBC

How AI video games can help reveal the mysteries of the human mind

AI integration in video games is helping to study the human mind. Games now use AI to create characters with detailed backstories, engaging players in unscripted conversations. Neuroscientists like Hugo Spiers utilize such games to research navigation and social behaviors, potentially offering insights into human cognition and relationships. MIT