Weekly AI News

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

Another jam-packed week in the world of AI, with plenty of big announcements. Nvidia unveiled stunning new processors and even built a robot simulator alongside Disney and Google (yep, you read that right!).

Meanwhile, Anthropic finally gave Claude real-time search and is working on a voice mode. Baidu’s not slacking off either: its Ernie 4.5 and Ernie X1 models aim to shake up the Chinese market and go head-to-head with DeepSeek.

Then, there’s Elon Musk snapping up a video-generation startup and Hollywood fuming over AI being trained on copyrighted works—time to grab the popcorn 🍿.

But the award for the most sensational development probably goes to humanoid robots: Figure plans to churn them out by the truckload in its new factory (100,000 a year!), while a Chinese robot is gearing up for a half marathon!

Between the copyright battle and the android invasion, it’s starting to feel like we’re watching a blockbuster storyline unfold. The difference is that it’s all happening… for real.

Happy reading!

🔥🤖 Nvidia unveils new AI chips, launches robot simulator with Google

Nvidia revealed a host of new software tools for robotics, including a simulator co-created with Disney and Google. These tools aim to boost the development of AI-driven robots powered by Nvidia’s chips. The company also previewed its upcoming Blackwell Ultra and Vera Rubin chips, promising doubled computing power each generation. In collaboration with Google and Disney Research, Nvidia introduced an open-source “physics engine” to help robots handle complex tasks more precisely.

⚙️🚀 Nvidia’s open-source reasoning models

Nvidia introduced Llama Nemotron, a family of open-source reasoning models designed to accelerate enterprise adoption of agentic AI. They come in three sizes—Nano, Super, and Ultra—and have already surpassed Llama 3.3 and DeepSeek V1 on several benchmarks. Each model can switch between deep reasoning and direct response based on the task at hand. Nvidia also announced an “AI-Q Blueprint” to help businesses integrate AI agents with existing systems.

🌐🔎 Claude gets real-time web search

Anthropic added web search capabilities to Claude, enabling real-time access to online information. Claude 3.7 Sonnet automatically decides when to search the internet and cites sources for easy fact-checking. The feature is now available to paid users in the U.S., with plans to expand internationally and to free tiers soon. This helps Claude catch up with other AI assistants that already have online search built-in.

🎤🗣️ Anthropic testing new voice features

Anthropic plans to introduce voice mode for Claude while focusing on enterprise markets. Chief Product Officer Mike Krieger mentioned upcoming features like calendar analysis and automated client reports, plus prototypes for voice experiences. Potential collaborations with Amazon and ElevenLabs could expedite the voice rollout. This move reflects Anthropic’s aim to stand out in the productivity and enterprise space.

💸⚡ China’s Baidu launches ultra-cheap AI models

Baidu rolled out Ernie 4.5 and Ernie X1 to compete with DeepSeek at half the cost. Ernie 4.5 claims to beat GPT-4o at just 1% of its price, while Ernie X1 matches DeepSeek R1 in reasoning performance. Both models will be accessible through Baidu’s cloud, with minimal token fees. Baidu hopes these releases will bolster its position in China’s fiercely competitive AI market.

⚖️📝 OpenAI & Google push for Fair Use in AI training

OpenAI and Google are lobbying for an expanded interpretation of “fair use” so they can train AI on copyrighted material without explicit permission. They argue it’s necessary for technological progress and doesn’t harm original creators. However, they face legal challenges from entities like The New York Times. Both companies have submitted proposals to U.S. policymakers to clarify fair use for AI.

🔊✨ OpenAI’s voice AI with personality boost

OpenAI introduced next-generation speech models for text-to-speech and speech-to-text, letting developers customise vocal styles with text prompts. The new gpt-4o-mini-tts can mimic various speaking tones—like a pirate or bedtime storyteller. Meanwhile, GPT-4o-transcribe shows state-of-the-art accuracy, surpassing Whisper. OpenAI also launched openai.fm, a demo platform to showcase these new capabilities.

🔗📂 OpenAI to start testing ChatGPT connectors for Google Drive and Slack

OpenAI will soon let ChatGPT Team subscribers connect the chatbot to Google Drive and Slack, enabling it to answer questions with context from those platforms. The data synced from these apps won’t be used for model training. OpenAI plans to extend support to more platforms like Microsoft SharePoint and Box, enhancing ChatGPT’s role as an enterprise-ready assistant.

⚡🔋 OpenAI’s Data Center Provider Lands Record Power Deal

Crusoe, part of OpenAI’s Stargate project, secured 4.5 gigawatts of power—enough to operate millions of Nvidia chips. Already building one of OpenAI’s largest data centers, Crusoe will resell capacity to major cloud providers and AI developers who are scrambling for server space. The deal with Engine No. 1 indicates an unprecedented scale for AI-focused infrastructure.

🚪🚶 OpenAI Post-Training Head Departs

Liam Fedus, OpenAI’s VP of research for post-training, is leaving the company to launch a startup focused on using AI to discover new materials. Fedus oversaw the models powering ChatGPT and OpenAI’s developer tools. OpenAI confirmed plans to invest in his new venture. His departure adds to a growing list of OpenAI executives venturing into their own startups.

⚖️🚫 Judge rejects Musk’s bid to block OpenAI’s evolution

A federal judge denied Elon Musk’s request for a preliminary injunction against OpenAI’s restructuring while expediting the trial. Court documents reveal Musk once contemplated merging OpenAI with Tesla as a for-profit entity, contradicting his current stance. With OpenAI eyeing a $40B SoftBank investment, this legal dispute could significantly impact the company’s path and broader AI developments.

🎥💡 Elon Musk’s xAI acquires video-generation startup

xAI has bought Hotshot, a text-to-video AI startup, as confirmed by CEO Aakash Sastry on X. Hotshot’s models (Hotshot-XL, Hotshot Act One, etc.) could help xAI develop “Grok Video” to compete with OpenAI’s Sora. Terms of the deal remain undisclosed, but some Hotshot team members are expected to join xAI.

🌐💰 xAI Joins Microsoft, Blackrock in Effort to Raise $100 Billion For AI Data Centers

Elon Musk’s xAI is part of a consortium including Microsoft, Blackrock, MGX, and Nvidia, aiming to fund a $100B initiative for AI data centers and power infrastructure. The partnership remains “non-exclusive,” implying multiple AI companies can benefit. MGX also supports Stargate, a huge data center venture involving OpenAI, Oracle, and SoftBank in the U.S. The goal: secure enough computing capacity to drive next-gen AI services.

🔮🤖 AI that can match humans at any task will be here in five to 10 years, Google DeepMind CEO says

Demis Hassabis believes we’ll reach artificial general intelligence (AGI) within a decade. Current AI systems are limited, but he foresees rapid advancements bridging the gap. While today’s models are still somewhat passive, he expects them to gain more human-like capabilities. Hassabis defines AGI as an intelligence matching the full spectrum of human cognitive abilities.

🗺️📝 Google’s NotebookLM added a new Interactive Mind Map feature

NotebookLM now lets users convert notes and source materials into visual mind maps. This interactive layer makes it easier to explore, connect ideas, and refine research. It’s designed to help people get a clearer overview of complex topics in a more intuitive way.

🖥️🤝 Google introduced Canvas in Gemini and added Audio Overviews

Gemini now features Canvas, a collaborative workspace where users can co-edit documents and code in real-time. Just click “Canvas” in the prompt bar to begin collaborating. Google also introduced Audio Overviews, letting Gemini summarise and present information audibly, facilitating efficient brainstorming or prototyping sessions.

📧✨ Google is improving Gmail’s search with AI

Gmail searches will now factor in your most-clicked emails and frequent contacts rather than just keywords and timestamps. This provides “most relevant” results at the top, though you can still switch back to the traditional chronological view. The feature is rolling out worldwide on web and mobile, aiming to help users find important emails faster.

🛰️🔥 Google’s AI Satellite to spot wildfires early

Google Research and Muon Space launched FireSat, an AI-driven satellite that detects wildfires as small as a classroom within minutes. Conventional satellites often miss smaller or early-stage fires. FireSat is the first of 50 satellites set to scan Earth’s surface every 20 minutes, helping firefighters respond quickly and improving our scientific understanding of wildfire patterns.

🧠⚗️ Microsoft teams with neuroscience AI startup Inait

Microsoft is partnering with Inait, a startup inspired by neuroscience, to develop AI that learns from real-world experiences rather than just pattern recognition. This brain-like approach aims to create more adaptive, generalisable AI systems.

📈🚀 Meta announced that its Llama open-source model officially reached 1B downloads

Llama soared to 1 billion downloads, up from 650 million in December 2024—underscoring the popularity of open-source AI initiatives. Researchers and developers embrace Llama’s flexibility, fueling innovation and collaboration on a massive scale.

🇫🇷💬 Meta deploys its AI chatbot in France

Meta AI, available in the U.S. for 18 months, is expanding to France and 40 other European countries. Users can ask questions and chat on WhatsApp, Instagram, and Facebook, free of charge. Meta hopes the chatbot’s international rollout will provide more useful, interactive experiences directly within its popular social apps.

⚡🔓 Mistral AI released Small 3.1

Mistral AI unveiled Small 3.1, a fast, open-source multimodal model featuring a 128k token context window. It outperforms Gemma 3 and GPT-4o Mini on several benchmarks. Mistral’s commitment to openness aims to empower developers with a versatile and high-performing model.

🏛️🌐 Hugging Face published a response to the White House’s AI Action Plan

In its “Open-Source AI Action Plan,” Hugging Face urged the U.S. government to favor open-source AI systems over commercial models. It outlined three pillars: strengthening open ecosystems, enabling efficient adoption, and promoting security standards. Hugging Face believes open-source AI rivals commercial offerings while providing economic, innovation, and cost advantages.

🔍💸 Perplexity in Talks to Raise at $18 Billion Valuation

Perplexity, an AI-powered search platform, seeks to raise between $500 million and $1 billion at a valuation of $18 billion—double its December figure. Led by CEO Aravind Srinivas, it nears $100 million in annual recurring revenue. Perplexity plans to diversify with advertising and developer APIs for AI search, bolstering its position among top-tier AI startups.

🧩🤖 Adobe’s army of enterprise AI agents

Adobe revealed a strategy centred on its new Experience Platform Agent Orchestrator, introducing ten specialised AI agents for tasks like content creation, site optimisation, and B2B management. A Brand Concierge bot personalises site chats as traffic from AI platforms soared 1,200%. Adobe Marketing Agent integrates with Microsoft 365 Copilot, and Adobe has alliances with AWS, SAP, and ServiceNow to ensure broad enterprise coverage.

💡🔧 Cursor added a MAX mode to use Claude 3.7 Sonnet in the app

Cursor’s new MAX mode taps into Claude 3.7 Sonnet with up to 200 tool calls and a 200k token context window. Priced at $0.05 per prompt or tool call, it offers unprecedented depth and flexibility for developers needing extended reasoning capabilities in their AI workflows.

🎞️🎨 Pika Labs released 16 new effects for its AI video platform

Pika Labs introduced 16 effects to morph images into various character videos. Creators can transform visuals in new ways—expanding the platform’s toolbox for imaginative video editing and storytelling through AI.

🔓🔊 Sesame open-sourced its Conversation Speech Model (CSM-1B)

Sesame, known for its ultra-realistic AI voice demos, has made its text-to-speech model CSM-1B publicly available. The open-source release encourages broader research and development of more lifelike, customisable voice applications.

🎙️🚀 Vogent AI launched voice agents that design and improve themselves

Vogent AI unveiled voice agents capable of self-improvement by learning from real failure cases without the need for intricate prompt engineering. Each misstep helps refine the models, suggesting ever-more effective and adaptive AI solutions over time.

🍏🔄 Apple Shakes Up AI Leadership After Delays

Apple tapped Mike Rockwell, head of the Vision Pro team, to steer Siri’s development away from AI chief John Giannandrea. CEO Tim Cook reportedly lost faith due to repeated delays and underwhelming features under the Apple Intelligence banner. Rockwell aims to revitalise Siri and other AI projects to meet Apple’s ambitious goals.

🗣️🤖 Zoom’s AI Companion goes agentic

Zoom upgraded its AI Companion with new agentic skills, enabling it to identify and carry out tasks across Zoom’s ecosystem. It can automatically detect action items in meetings, handle follow-ups, and create documents with minimal user input. A $12/month add-on also includes personal AI coaches, AI avatars, and more, marking Zoom’s pivot toward fully automated collaboration.

🎮🔨 Roblox releases open-source 3D generation AI

Roblox introduced Cube 3D, which creates 3D objects and scenes from text prompts using native 3D data. It generates full functional assets—like “/generate motorcycle”—with future capabilities planned for image inputs. Roblox also improved its Studio suite for real-time collaboration and monetisation, aiming to streamline game development for its 85M+ daily users.

🏭💡 SoftBank buys Sharp plant for OpenAI project in Japan

SoftBank is investing $676 million to acquire Sharp’s old LCD factory in Osaka and transform it into an AI data center for “Cristal Intelligence,” in partnership with OpenAI. The facility will have high power capacity to train advanced models and plans to go operational by 2026, marking SoftBank’s third data center in Japan.

🚩🖼️ China will enforce clear flagging of all AI-generated content starting in September

From September 1, 2025, any AI-generated text, images, audio, video, or virtual scenes must be visibly and metadata-labeled as such. App store providers must ensure that hosted AI apps follow the new rules. While users can request unlabeled content, the provider must log this and remind them of the regulations.

⚖️🚫 US appeals court rejects copyrights for AI-generated art

A federal appeals court upheld that works solely generated by AI, with no human input, cannot be copyrighted under U.S. law. Stephen Thaler’s “A Recent Entrance to Paradise” was cited as lacking a human author. The ruling confirms prior positions by the U.S. Copyright Office but still allows copyrights for AI-assisted pieces with significant human involvement.

📈🔮 Study: AI capabilities following ‘Moore’s Law’

A METR study shows the time span of tasks AI agents handle autonomously has doubled every seven months since 2019. Models like 3.7 Sonnet can successfully manage tasks equivalent to 59 minutes of human work with 50% reliability. If the trend persists, AI could tackle months-long human-equivalent projects by 2030, underscoring the potential for exponential leaps.

😂🖼️ AI Could Beat Humans at Meme Creation, Study Finds

According to the European study, “One Does Not Simply Meme Alone,” AI-generated memes scored highest on humour, creativity, and shareability. Humans still produced the funniest individual memes, while human AI teams created the most original, viral content. Researchers say AI’s “crowd-pleasing” humour arises from massive training datasets and can spark even more ideas in collaboration with humans.

💼📉 Only 35% of employees using AI at work think their company’s deployments are successful, vs. 75% of execs

Research by Writer highlights a significant disconnect between corporate leaders and frontline workers on AI effectiveness. Interestingly, 35% of employees pay out of pocket for AI tools rather than using company-provided ones. This gap suggests many leaders overestimate how well AI is serving their teams.

🩺💊 Harvard team creates AI agent for personalised medicine

A joint Harvard-MIT project unveiled TxAgent, which uses multi-step reasoning and real-time medical data retrieval to recommend personalised treatments. It checks for drug interactions, genetic factors, and comorbidities, refining its suggestions iteratively. With 211 specialised tools and trusted data sources, TxAgent aims to deliver safer, more tailored care quickly.

🤖🏗️ Humanoid robot maker Figure announced the launch of BotQ

Figure is opening a new manufacturing facility, BotQ, capable of producing 12,000 humanoid robots annually. Plans are in place to scale this to 100,000 per year to meet rising demand in industrial and logistics applications. The move could solidify Figure’s position in the growing humanoid robotics market.

🏃‍♂️⚙️ Chinese robot prepares for half marathon

An enhanced Chinese humanoid robot is being tested for a half marathon, featuring longer legs, sturdier hips, and custom shock-absorbing “feet.” With a top speed of 12 km/h, it will likely run slower for stability. The event, planned for April 13, will include participants from robotics companies, research institutes, clubs, and universities worldwide.

🎬⚖️ Hollywood against AI copyright proposals

Over 400 Hollywood creatives—including Ben Stiller, Paul McCartney, and Cate Blanchett—urged the Trump administration to reject AI fair use expansions proposed by OpenAI and Google. They argue that an “unrestricted fair use” would allow “free exploitation” of copyrighted works, calling for proper licensing negotiations. Meanwhile, AI models worldwide continue ingesting content, making this clash both symbolic and significant.

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