? AI Flash: Mar 07, 2025

Twice a week, we cut through the noise to bring you the most important AI breakthroughs, research, and industry moves — sharp, concise, and straight to the point.?

In Today's AI Flash?

Top AI Headlines?

Еric Schmidt Warns of AI Arms Race, Proposes "Mutual Assured AI Malfunction"?

Former Google CEO Eric Schmidt, along with AI leaders, is warning against a "Manhattan Project"-style push for superintelligent AI, arguing it could escalate into a dangerous global arms race.?

• Schmidt and his co-authors advocate for deterrence, transparency, and international AI cooperation rather than reckless acceleration.?
• They propose Mutual Assured AI Malfunction (MAIM)—a concept mirroring Cold War-era nuclear deterrence to prevent AI dominance by any single power.?

• The warning comes as Trump’s administration commits $500B to AI infrastructure, intensifying U.S.-China competition.?

Business Takeaway: AI governance is becoming a high-stakes geopolitical issue. Businesses investing in AI need to monitor emerging regulations to navigate compliance, security risks, and potential restrictions on AI development.?

?? Full Story?

OpenAI Reportedly Plans to Charge Up to $20,000 a Month for Specialized AI Agents?

OpenAI is rumored to be launching high-end, specialized AI agents with pricing reaching $20,000 per month for premium tiers.?

• The company is developing agent products tailored to specific industries, including sales lead sorting, software development, and PhD-level research.?

• SoftBank has reportedly committed $3B to OpenAI’s agent offerings as the company seeks profitability after a $5B loss last year.?

• The highest-tier $20,000/month agent is aimed at handling complex research tasks, pushing AI closer to true expert-level reasoning.?

Business Takeaway: AI services are entering premium enterprise territory, signaling that AI tools delivering real business value will justify steep pricing. Companies must assess ROI before integrating costly AI solutions.?

?? Full Story?

?

Google Co-Founder Larry Page Reportedly Launches New AI Startup?

Larry Page is reportedly working on a stealth AI venture, Dynatomics, focused on AI-driven product manufacturing.?

• The startup is developing AI models that create highly optimized designs for physical products, integrating AI into factory workflows.?

• Chris Anderson, former CTO of Kittyhawk, is leading the company’s engineering efforts.?

• The move aligns with a broader trend of AI-driven manufacturing innovation, similar to companies like Orbital Materials and PhysicsX.?

Business Takeaway: AI-powered design and automation are transforming manufacturing. Businesses should explore AI-driven optimizations for supply chains, materials engineering, and production efficiency.?

?? Full Story?

?

China’s Manus AI Unveils "Fully Autonomous" AI Agent?

A Chinese startup has introduced Manus AI, claiming it to be the first AI agent capable of fully autonomous task execution, achieving state-of-the-art results on agentic benchmarks.?

• Manus AI can screen resumes, conduct property research, browse the web, code, and create visuals—operating independently on its own computer instance.?

• It reportedly outperformed ChatGPT and Gemini on GAIA, a benchmark for AI task automation.?

• The agent remains invite-only, with plans to open-source the underlying models later this year.?

Business Takeaway: Autonomous AI agents are moving beyond simple automation and approaching real-world workflow independence. Businesses relying on repetitive tasks should start considering AI-driven agentic automation.?

?? Full Story?

Tavus Unveils Emotionally Intelligent AI for Conversational Video?

Tavus has introduced the next evolution of Conversational Video Interface (CVI)—bringing AI-generated avatars closer to human-like interactions.?

• Phoenix-3 enables full-face animation, making avatars express emotions naturally in real-time.?

• Raven-0 acts as the avatar’s "eyes," analyzing body language and facial expressions to respond appropriately.?

• Sparrow-0 improves conversation flow, eliminating awkward pauses and robotic timing.?

Business Takeaway: AI-generated avatars are evolving from scripted to emotionally intelligent interactions. This opens new possibilities for AI-powered customer service, sales enablement, and digital assistants.?

?? Full Story?

AI Papers Highlights

ThunderMLA: A Faster, Fused Multi-Headed Attention Kernel?

Key Idea: Stanford’s Hazy Research team introduced ThunderMLA, a "megakernel" approach that makes LLM attention decoding 30% faster than existing implementations.?

Implications: By fusing kernels into a single, highly optimized unit, ThunderMLA reduces memory overhead and improves efficiency in variable-length sequence processing.?

Takeaway for Leaders: Businesses using LLMs at scale should explore custom kernel optimizations like ThunderMLA to lower inference costs and boost performance.?

?? Link to Paper?

?

Maximizing LLM ROI: The Hidden Costs of AI Deployment?

Key Idea: A new report from Turing AI reveals that LLM costs are often inflated by poor training, misaligned evaluation metrics, and inefficient optimizations.?

Implications: Companies investing in AI often fail to optimize models for real-world use, leading to increased costs and lower ROI.?

Takeaway for Leaders: AI-driven enterprises should focus on evaluating LLM performance beyond just raw power, prioritizing alignment, fine-tuning, and cost-effective inference strategies.?

?? Link to Paper?

?? Data and Analytics Tip

How to Optimize Power BI Data Models for Large Datasets

Handling massive datasets in Power BI requires a good star schema but it also demands advanced optimizations to improve refresh times, reduce model size, and boost query performance. Our latest guide covers:?

? Reducing model size: Remove unnecessary rows/columns and optimize data storage.?

? Choosing the right storage mode: Import, DirectQuery, Dual, or Direct Lake — what works best for your dataset??

? Improving refresh efficiency: Use incremental refresh and aggregations for faster data updates.?

?? Read the Full Guide?