20 Jul 2025

AI News Digest

🤖 AI-curated 8 stories

Today's Summary

OpenAI’s new GPT-5 Multimodal is out in public beta, letting you interact with text, images, audio, and video all at once, which feels like a glimpse into a much more intuitive future for AI. Meanwhile, Amazon picked up Bee, a startup with a knack for AI wearables, to beef up its ambient intelligence game—think voice-recording bracelets that help you stay on top of your to-do list. And if you’re curious about what’s hot in AI tools this year, TechRadar just dropped a massive review of over 70 options, so there’s plenty to explore if you’re looking to bring a little more AI into your workflow.

Stories

OpenAI Releases GPT-5 Multimodal Public Beta

OpenAI has debuted the public beta of GPT-5 Multimodal, a large language model capable of processing text, images, audio, and video simultaneously. This advancement represents a major step in multi-modal AI, making human–computer interaction more intuitive and widening potential applications—from customer support to accessibility tools.

Amazon Acquires AI Wearables Startup Bee to Expand Ambient Intelligence Offering

Amazon has acquired Bee, an AI wearables startup known for its voice-recording bracelet and Apple Watch app. Bee’s product captures ambient conversations to generate reminders and to-do lists, functioning as a low-cost AI assistant. This acquisition aims to enhance Amazon's ambient intelligence offerings.

BLIP3-o: Open-Sourced Multimodal Models Achieve State-of-the-Art Performance

Researchers have introduced BLIP3-o, a suite of fully open unified multimodal models designed for both image understanding and generation. By integrating a diffusion transformer for generating CLIP image features and implementing a sequential pretraining strategy, BLIP3-o enhances training efficiency and generative quality. The models have achieved state-of-the-art results across various benchmarks and are fully open-sourced, including code, model weights, and datasets. This development signifies a significant advancement in unified multimodal modeling, offering practical benefits and supporting reproducibility in AI research.

AlphaGenome: DeepMind's Unified Model Advances Genomic Prediction

Google DeepMind has introduced AlphaGenome, a deep learning model capable of predicting thousands of functional genomic tracks directly from DNA sequences at single base-pair resolution. By unifying multimodal prediction, long-range sequence context, and high resolution in a single framework, AlphaGenome matches or exceeds the performance of specialized models across various genomic tasks. This breakthrough enhances the interpretation of non-coding variants in disease and supports large-scale genome analysis, marking a significant step forward in genomic research.

Read more → Your guide to AI: July 2025

Anthropic Seeks to Double Valuation to Over $150 Billion in New Funding Round

AI startup Anthropic is in early discussions to raise a new funding round that could more than double its valuation to over $150 billion, up from its current $61.5 billion. The company, which competes with OpenAI in developing advanced AI models, is expected to raise at least $3 billion, potentially up to $5 billion. Interest has emerged from several large Middle Eastern investors, including Abu Dhabi-based AI fund MGX. Although Anthropic has been cautious about accepting funding from the region due to ethical concerns, prior secondary share purchases linked to MGX have occurred. Backed by Google and Amazon — the latter with a current $8 billion commitment and potential for further investment — Anthropic is ramping up its business-focused subscription revenue, now exceeding $4 billion annually. However, the company, like its rivals OpenAI and xAI, remains unprofitable due to high computing costs and fierce competition for AI talent. This funding round positions Anthropic as a major contender in the AI space, where OpenAI continues to lead, with ChatGPT surpassing 500 million weekly users and a $300 billion valuation.

Elon Musk's xAI Seeks $12 Billion in Debt Financing for AI Expansion

Elon Musk's AI startup, xAI, is reportedly seeking to raise up to $12 billion in debt financing to support its expansion, according to the Wall Street Journal. Valor Equity Partners, closely tied to Musk through its founder Antonio Gracias, is in discussions with lenders to secure the funding. The capital will be used to acquire a large number of advanced Nvidia chips that will be leased to xAI for powering and training its AI chatbot, Grok, in a massive data center. Some lenders are proposing a three-year loan term with borrowing caps to mitigate risk. xAI is currently training Grok using a supercluster of 230,000 GPUs, including Nvidia GB200 chips, with plans to add another supercluster using 550,000 GB200 and GB300 chips. Despite reports that xAI may spend around $13 billion in 2025 and could be valued between $170 billion and $200 billion, Musk has publicly claimed that the company is not actively seeking funding and has adequate capital.

Comprehensive Review of Over 70 AI Tools in 2025

TechRadar has published an extensive guide evaluating more than 70 AI tools across various domains, including chatbots, image generation, virtual assistants, and educational platforms. The review highlights leading tools such as ChatGPT-4o, Google Gemini, and Grok AI for conversational AI, and Adobe Firefly and Wix for AI-powered design and website building. This resource serves as a valuable reference for professionals seeking to integrate AI into their workflows.

Design+Code Launches 'Create Your Dream Apps with Cursor and Claude AI' Course

Design+Code has introduced a new course titled 'Create Your Dream Apps with Cursor and Claude AI,' aimed at teaching developers how to build web applications using AI tools. The course covers frontend development with React, backend integration with Firebase, and payment handling with Stripe. It also explores advanced AI tools like Claude Artifacts and Galileo AI, providing a comprehensive learning experience for aspiring app developers.