Matthew Berman
April 07, 2025
Good morning, it’s Monday. China’s latest move to mirror Trump’s trade tactics may sting more at home than abroad—and in AI, it’s not just about who builds the smartest models anymore. It’s about who’s actually using them (spoiler: China’s catching up fast).
Plus: 🦙 Llama 4 has arrived—and it’s bringing the whole herd.
Read on!
SandboxAQ, a quantum-tech startup spun off from Google’s parent Alphabet in 2022, has raised an additional $150 million from investors including Google and NVIDIA. The infusion brings the company’s Series E round to a total of $450 million and lifts its valuation to $5.75 billion, with total funding to date nearing $950 million. → Continue reading here.
🦙 Meta Unveils Llama 4 AI Models
Meta has introduced the Llama 4 series, comprising Scout, Maverick, and the forthcoming Behemoth models, marking a significant leap in AI's ability to process text, images, and video. Scout, a compact model, boasts a 10-million-token context window and outperforms rivals like Google's Gemma 3 in efficiency. Maverick, with 400 billion parameters, matches OpenAI's GPT-4o in coding and reasoning tasks while utilizing fewer resources.
🚀 OpenAI Revives o3, Delays GPT-5 Launch
OpenAI is reversing course on its shelved o3 model, now set to launch alongside a lighter o4-mini version within weeks. CEO Sam Altman says the delay in GPT-5 stems from integration challenges—but also a chance to improve it beyond expectations. GPT-5, now expected in a few months, will feature tiered intelligence access and unify tools like search and voice.
🖼️ Midjourney Launches V7
Midjourney has debuted V7, its first new image model in nearly a year, featuring a “totally different architecture” and default personalization based on user-rated images. V7 improves prompt coherence, texture quality, and object accuracy, and introduces Draft Mode for 10x faster, lower-cost rendering. It comes in Turbo and Relax modes, though some features like upscaling aren’t yet available.
🎮 Microsoft Rebuilds Quake II with AI: Using its Muse AI model, Microsoft recreated scenes from the classic shooter as a tech demo — showing how Copilot-like tools could power future game design.
🏷️ Google Unveils Gemini 2.5 Pro Pricing: At up to $15 per million tokens, it’s Google’s most expensive AI yet—beating out Gemini 2.0 Flash and even OpenAI’s o3-mini. Still, devs are calling it worth it for top-tier reasoning, coding, and math.
🦠AI Maps How Cells Fight Disease: Tel Aviv researchers built scNET, an AI tool that filters cellular “noise” by merging gene data with a protein social network, revealing hidden immune responses to treatments like cancer drugs.
🔲 Rapidus Courts Apple, Google for Chips: Japan’s Rapidus is in talks with Apple, Google, and others to mass-produce advanced AI chips by 2027, aiming to challenge TSMC. Backed by $1.37B federal funding amid rising tensions.
🛍️ Amazon Tests AI That Buys for You: “Buy for Me” lets users purchase from third-party sites without leaving the app—Amazon’s AI fills in your details and checks out for you, aiming to outdo rivals like Google and OpenAI.
👩‍💻 Copilot Levels Up with Web Actions: Microsoft’s Copilot can now book trips, track deals, read your screen, and remember your preferences—pushing deeper into AI assistant territory with new tools rivaling ChatGPT and Gemini.
🏗️ Feds Open Labs for AI Data Centers: The U.S. Energy Department is inviting tech firms to build AI data hubs on 16 federal sites—including Los Alamos—offering fast-track access to permits in a bid to supercharge AI development.
𖦹 AI Shakes Up Fingerprint Science: A Columbia-led study found that different fingers from the same person can produce similar prints, challenging a century-old forensic belief—and sparking debate over how law enforcement IDs suspects.
Forward Future is powered by readers like you. By upgrading to Premium, you’ll get exclusive insights, members-only content, and directly support our work.
Here’s what you get with Premium:
🔍 Exclusive Deep-Dive Content
🧠“I Will Teach You How to AI” Series
🎤 Insider Interviews
🧰 AI Tool Library (Coming Soon!)
🏗️ AI Model Hub (Coming Soon!)
The Recap: The Economist compares AI adoption between the U.S. and China, finding that while the U.S. still leads in infrastructure and enterprise software, China is rapidly catching up in everyday use across government, consumer tech, and manufacturing. Despite assumptions that China lags in implementation, new data shows it may already be ahead in certain areas. The analysis highlights that widespread, real-world deployment—not just model innovation—will determine global AI leadership.
Highlights:
While OpenAI leads in model development, Chinese companies like DeepSeek offer near-equivalent models at significantly lower cost.
China rose 15 spots in the World Intellectual Property Organization’s diffusion ranking since 2020, now at 32nd globally.
IBM reports that 50% of Chinese firms use AI, versus one-third in the U.S.; McKinsey finds more Chinese workers use AI on the job.
State adoption in China is extensive, including AI for public services, while consumers benefit from free or cheap AI apps from firms like ByteDance and Tencent.
Manufacturing accounts for 43% of China’s AI-focused VC funding, compared to just 3% in the U.S.; China now commands nearly 30% of global manufacturing output.
Despite U.S. tech giants planning to invest $300B in AI this year, Chinese firms are stretching smaller budgets further, aided by open-source models and lower cloud costs.
American firms still dominate in infrastructure: the U.S. has 10 times more data centers and quadruple the AI server spending compared to China.
Forward Future Takeaways:
The global AI race is shifting from who builds the best models to who puts them to work most effectively—and China is showing surprising strength in real-world deployment. As AI becomes embedded in everything from government services to industrial robots, diffusion—not innovation—may be the more decisive metric of leadership. This raises a crucial question: Can American firms and regulators keep pace with the speed and scale of China's grassroots AI adoption? → Read the full article here.
Companies generate a ton of valuable information in Slack threads, Notion docs, and meeting notes—but it often gets buried. Turning this into a structured, AI-powered Q&A system helps teams instantly surface key knowledge, reduce repeated questions, and onboard faster. → Continue reading here.
OpenAI's Paperbench shows AI agents can replicate cutting-edge ML research from scratch—hinting at self-improving AI. Claude 3.5 leads, but agentic scaffolds are key to future progress. Get the full scoop in Matt’s latest video! 👇
🤖 Zapier MCP: Lets AI assistants run 30K+ actions across 7K+ apps—no APIs needed. Automate tasks like messaging, scheduling, and data updates.
đź“ť Mem AI: An AI-powered note-taking app that organizes your notes automatically, offers smart search, and provides AI-assisted writing.
✍️ LanguageTool: AI Writing Assistant that checks grammar, spelling, and style in 30+ languages—seamlessly integrates across browsers, apps, and documents.
What did you think of today's newsletter? |
❤️ Love Forward Future? Spread the word & earn rewards! Share your unique referral link with friends and colleagues to unlock exclusive Forward Future perks! 👉 Get your link here.
📢 Want to advertise with Forward Future? Reach 450K+ AI enthusiasts, tech leaders, and decision-makers. Let’s talk—just reply to this email.
🛰️ Want more Forward Future? Follow us on X for quick updates, subscribe to our YouTube for deep dives, or add us to your RSS Feed for seamless reading.
Thanks for reading today’s newsletter—see you next time!
The Forward Future Team
🧑‍🚀 🧑‍🚀 🧑‍🚀 🧑‍🚀
Reply