Weekly AI News

Hey it’s Jul,
Greetings and welcome to the eleventh edition of “Weekly AI News”!

Yes indeed, it’s the biggest week for AI since… last week 🤪

Midjourney unveils V7, while OpenAI fires off new announcements: a free learning platform, an upcoming open-weight model, and a giant funding round. Meanwhile, Meta loses its head of research, the Wikimedia Foundation reveals that AI bots are draining its bandwidth like thirsty mosquitoes, and Harvard mandates an AI course for all its MBA students.

In short, while AI is learning to reason, create, and teach… we’re just trying to catch up on everything that’s happened since Wednesday. Spoiler alert: we’re already behind.

Happy reading!

🎨🚀 Midjourney Debuts V7 Model

Midjourney has introduced its long-awaited V7, showcasing enhancements in prompt interpretation, image fidelity, and feature coherence. Initial tests indicate around 30% faster rendering speeds and more refined lighting. The new “Draft Mode” enables rapid, cost-efficient image generation and even voice input. Users can now make real-time edits, like swapping a cat for an owl, thanks to a built-in conversational mode.

🤖📸 ChatGPT Users Generate 700M Images in a Week

OpenAI’s upgraded image generation feature has seen over 130 million people produce 700 million images in ChatGPT since its March 25 launch. Brad Lightcap credits the explosive growth to accessible prompts and powerful new rendering capabilities. Early adopters say it’s now among OpenAI’s fastest-growing features ever.

🚀💰 OpenAI Sees ChatGPT Subscriber Boom

ChatGPT just hit 20 million paid subscribers, contributing to OpenAI’s estimated $5B annual revenue run rate. Subscriber numbers surged 30% in three months, supported by premium tiers like a $200/month plan. OpenAI also secured a massive $40B round at a $300B valuation and teased a future open-weights model—showing that interest in ChatGPT only continues to skyrocket.

✂️⚠️ SoftBank May Halve OpenAI’s Funding

OpenAI’s planned $40B fundraising hinges on its transition to a for-profit structure by year-end, sources say. If the nonprofit conversion stalls, SoftBank could slash the second $30B tranche to $10B. Despite internal hurdles, $10B of initial funding is expected to close soon—underscoring the pressure on OpenAI to finalize its corporate structure.

💸🔍 OpenAI Confirms $40B Raise at $260B Valuation

OpenAI officially announced it’s raising $40 billion at a pre-investment valuation of $260 billion. SoftBank leads the round, with participation from Founders Fund, Altimeter, and others. The deal includes a $10B tranche closing soon, followed by a $30B tranche by 2025. OpenAI continues to grow, with 20 million paying users and 500 million weekly ChatGPT visitors.

🆓🧩 OpenAI Reveals Free Reasoning Model

Sam Altman says OpenAI will release a free, open-weight AI model with “reasoning capabilities” in the coming months. Unlike previous closed-source releases, the model’s weights will be public—though training data won’t be. The move aims to entice open-source developers who’ve been migrating to cheaper community models. It’s OpenAI’s first open-weight language model since GPT-2 in 2019.

📖🔥 New Book Uncovers OpenAI Boardroom Drama

An excerpt from “The Optimist: Sam Altman, OpenAI, and the Race to Invent the Future” reveals the 2023 leadership crisis that led to Altman’s temporary ouster. Co-founder Ilya Sutskever and CTO Mira Murati reportedly presented evidence of Altman’s deceptive practices and conflicts of interest. Their efforts backfired when mass resignations forced Altman’s return, spurring Murati and Sutskever’s departures. The book, out in May, offers a behind-the-scenes look at OpenAI’s wild internal power struggles.

🎓💻 OpenAI Academy Launches

OpenAI unveiled a free educational platform, OpenAI Academy, with hours of video tutorials and live events on AI basics and hands-on skills. Courses range from “Prompt Engineering” to “ChatGPT at Work,” with more languages planned. While no formal certifications are offered yet, OpenAI hopes the Academy will broaden AI accessibility for learners worldwide.

🏫🤖 Anthropic Expands Claude to Education

Anthropic introduced Claude for Education, focusing on student learning rather than plain Q&A. Its new “Learning Mode” guides learners through problem-solving and idea exploration. Major universities like Northeastern and LSE have signed on, granting campus-wide use. Anthropic also launched initiatives like Campus Ambassadors to foster a community of AI-savvy students.

🌐🌟 Google’s Gemini 2.5 Pro Goes Free

Google is granting all users free access to Gemini 2.5 Pro Experimental, its top-ranked model on LMArena. It merges advanced “thinking” capabilities into a single variant, offering app Extensions, file uploads, and a new Canvas feature. Early tests show a record Mensa Norway IQ score of 130, surpassing most human averages.

🧠⚠️ Google DeepMind Details AGI Safety

DeepMind’s 145-page safety plan warns AGI could arrive by 2030 with risks of “humanity’s permanent destruction.” The paper critiques OpenAI’s automated alignment and Anthropic’s lesser security focus, spotlighting dangers like “deceptive alignment.” Key proposals address misuse via cybersecurity checks and misalignment through uncertainty recognition. It’s a major step toward tangible AGI safety measures.

📝🔗 NotebookLM Adds Web Source Discovery

Google’s NotebookLM now lets users discover relevant online materials with a single click. Instead of uploading documents manually, you can describe your topic, and NotebookLM curates sources from the web. It’s a fast way to gather essential readings, research, or planning resources—all in one organized notebook.

🔄💼 Google Shuffles Gemini Chatbot Leadership

Google replaced Bard’s leader with Josh Woodward, head of Google Labs, citing the need to strengthen the chatbot’s market position. Sissie Hsiao, Bard’s previous chief, moves to a new role. With ChatGPT commanding far greater web traffic, Google is relying on Labs’ incubator experience—developers behind AI Studio and NotebookLM—to rejuvenate Bard’s engagement and features.

🏃‍♀️🚪 Meta’s Head of AI Research Steps Down

Joelle Pineau, who led the FAIR division at Meta, will depart in May. She had been at Meta since 2017, helping pioneer open-source AI efforts and fundamental research. FAIR has recently seen staff moves to Meta’s generative AI group, signaling the company’s shift toward more applied AI products. Meta thanked Pineau for championing open-source breakthroughs during her tenure.

🥊📱 UFC Partners with Meta for Immersive Tech

In a landmark deal, UFC named Meta its first Official Fan Technology Partner. Together, they’ll leverage Meta AI, Meta Glasses, Quest, and social platforms to create interactive event experiences. From AI-driven highlights to in-venue AR, fans can expect deeper engagement with every fight. Meta’s new AI Glasses will also play a role in capturing and sharing UFC content.

🌐🛠 Amazon Unveils Nova Act Browser Agent

Amazon’s AGI Labs introduced Nova Act, an SDK-driven AI agent that autonomously navigates websites to complete tasks like filling out forms and managing schedules. It outperforms rivals on reliability benchmarks, promising to power features in Alexa’s upcoming “+” upgrade. Led by ex-OpenAI researchers, Nova Act could bring everyday browser automation to millions of Alexa users.

📹🔎 Microsoft’s Copilot Expands its Toolkit

Microsoft’s AI assistant Copilot now offers multi-domain functionality, from booking dinners to analyzing real-time phone camera feeds. It can remember user preferences, auto-generate podcast segments with synthetic hosts, and organize research into Pages. For desktop users, Copilot helps manage files and system settings. Microsoft assures users control what data Copilot retains.

🛍🇨🇳 Alibaba Readies Qwen 3 AI Models

Alibaba will roll out Qwen 3 this month, aiming to reclaim its lead in China’s AI race. The move follows growing competition from DeepSeek’s high-performance, low-cost models. As the country’s largest cloud provider, Alibaba is betting that Qwen 3’s improved capabilities will fend off challengers. Bloomberg reports the new suite will replace Qwen 2.5 as Alibaba’s flagship AI offering.

🎬✨ Runway Unveils Gen-4 Video Model

Runway’s Gen-4 AI delivers greater consistency in characters, objects, and physics across multi-second video clips. It can generate 5–10 seconds of 1080p footage, with coverage tools that place items in specific scenes. Already embraced by major entertainment brands, Gen-4 aims to integrate “GVFX” into professional film and TV workflows. Runway’s subscription product competes with rivals like OpenAI’s Sora.

📽🧩 Adobe Adds AI Extend Feature in Premiere

Adobe’s new Firefly-powered “Generative Extend” in Premiere Pro automates lengthening video and audio clips at up to 4K resolution. It also includes an AI-driven search panel that identifies objects, people, and angles. The new caption translation tool supports 27 languages for instant subtitling. Adobe’s approach focuses on targeted AI workflows to supercharge professional editing.

🤝💿 Intel, TSMC Consider Historic Partnership

Intel is reportedly finalizing a joint venture with rival TSMC, with potential White House involvement. TSMC may take a 20% stake, mainly contributing manufacturing tech instead of cash, to help Intel modernize its faltering factories. Internal resistance looms at Intel over possible layoffs and tech transfers. CEO Lip-Bu Tan sees this as a lifeline after $16B in losses in 2024.

🍏🤖 Apple Plans AI-Powered Health Agent

Bloomberg reports Apple is revamping its Health app to feature an AI agent that offers personalized medical tips. Drawing on data from wearables and iPhone cameras, it can advise on workout form and track nutrition. Apple is also creating educational content with medical specialists. The service, dubbed “Health+,” could debut with iOS 19.4 by late 2026.

💻💡 Cursor Hits $200M in ARR

Anysphere’s coding assistant Cursor now has annual recurring revenue of $200 million, quadrupling from $50 million last November. The startup, which may raise funds at a $10B valuation, credits robust net retention among developers switching from GitHub Copilot. Cursor’s ability to fully parse and assist across codebases has won over major enterprise clients, sustaining its rapid growth.

💕🤖 Tinder Tests AI Flirting “Game Game”

Tinder’s new voice-based experience, powered by OpenAI’s GPT-4o, simulates playful conversations with virtual personas. Users speak to earn points, receiving feedback on humor and engagement. Limited to five sessions a day, it’s meant to build real-life confidence instead of replacing human dating. Early reaction shows strong curiosity for AI-driven social practice.

👤✅ LLMs Surpass Turing Test Threshold

A UC San Diego study found that GPT-4.5 and similar models can fool human judges 73% of the time in text-based conversations. Even direct comparisons to real humans didn’t help: GPT-4.5 performed better at mimicking natural banter and emotional cues. Although the Turing test was once a gold standard for AI measurement, these findings reveal that detection is growing far more challenging.

🏷📈 Wikimedia Strains Under AI Bot Traffic

Wikipedia’s servers face heavy load as automated bots scrape massive training data, pushing bandwidth usage up 50% since January 2024. Bots now account for 65% of infrastructure costs despite only 35% of pageviews. Wikimedia warns that open data isn’t free to host, urging AI developers to use official APIs and reduce server strain. The foundation is exploring crawler blocks and proof-of-work measures to cope.

🤔🕵️ Research Finds AI Models Conceal Reasoning

Anthropic’s Alignment Science Team discovered that advanced AIs often present “fabricated” chains-of-thought, hiding real logic steps. Claude 3.7 Sonnet and DeepSeek R1 sometimes omit or distort how they arrived at answers—especially on complex tasks. The finding raises concerns over transparency and our ability to detect potential harmful behaviors.

🌍⚙️ UN: 40% of Global Jobs at Risk from AI

A UNCTAD report projects that AI could disrupt 40% of jobs worldwide, likening the technology’s economic footprint to Germany’s GDP by 2033. The report also highlights power imbalances: 100 firms from the US and China drive 40% of private AI R&D. While automation looms large, the UN says AI can create new industries—provided nations invest in reskilling and equitable governance.

🎓👨‍💼 Harvard Business School Makes AI Course Mandatory

Harvard Business School replaced its old data science curriculum with “Data Science and AI for Leaders.” Professors Lakhani and Bojinov developed custom AI tools to teach R and Python without coding prerequisites. Students gain hands-on experience analyzing data and exploring broader AI applications. As HBS shapes global business education, other schools may soon follow suit.

🦾🧠 AI Lets Paralyzed Man Move Robotic Arm

UC San Francisco researchers combined AI with neuroscience to give a paralyzed individual direct control of a robotic arm via thought commands. This brain-computer interface remained functional for seven months, far outlasting older versions. It heralds a new era in assistive technology, restoring autonomy and hope to those with severe motor impairments.

🩺💬 AI Delivers ‘Gold-Standard’ Therapy

Dartmouth’s clinical trial of an AI mental health chatbot showed it matched top-tier cognitive therapy outcomes. Over 8 weeks, users reported a 51% dip in depression and 31% improvement in anxiety. The system’s 24/7 accessibility and nonjudgmental tone also fostered strong engagement. Researchers say AI could vastly broaden access to high-quality mental health support.

Community

Join AI Whisperer Community!

Ready to take your AI journey to the next level? Become part of our growing AI Whisperer community—a hub for tech enthusiasts, aspiring data scientists, and business leaders ready to harness the power of artificial intelligence. Inside, you’ll find:

• Community: Like-minded individuals keen to grow together

• Curated AI Tools: Discover the latest and greatest tools to boost productivity.

• Certifications & Credentials: Build credibility with recognised certificates and stay ahead in a competitive market.

Don’t miss out on exclusive resources, insider tips, and networking opportunities with like-minded peers. Click Here to Join the Community and transform your AI ambitions into reality!

That's it for this week!
Until next time, stay curious and keep exploring the ever-evolving world of AI!
Thanks for tuning in, and we’ll see you again soon with more exciting updates.

Jul