Good morning and thank you for joining us again!

Welcome to this daily edition of Beginners in AI, where we explore the latest trends, tools, and news in the world of AI and the tech that surrounds it. Like all editions, this is human curated and edited, and published with the intention of making AI news and technology more accessible to everyone.

THE FRONT PAGE

Morgan Stanley: AI to Cut 200,000 EU Banking Jobs by 2030

TLDR: Morgan Stanley forecasts European banks will cut 200,000 jobs by 2030 as AI takes over back-office work, but one top banker warns the industry is "storing up a big problem."

The Story:

A Morgan Stanley analysis of 35 European banks projects that AI and branch closures will eliminate roughly 10% of the continent's banking workforce, about 212,000 positions, by 2030. The cuts will hit hardest in "central services": back-office operations, risk management, and compliance, where banks say they're expecting 30% efficiency gains from AI. Dutch lender ABN Amro has already announced plans to slash 5,200 jobs, nearly a quarter of its workforce, by 2028, with AI handling routine tasks in customer service and anti-money laundering. Société Générale's CEO warned in March that "nothing is sacred" in his cost-cutting campaign, while Goldman Sachs launched its "OneGS 3.0" initiative targeting AI automation from client onboarding to regulatory reporting.

Its Significance:

Banking is becoming a preview of what AI-driven workforce restructuring looks like across white-collar industries. The roles being cut aren't factory jobs or manual labor. They're compliance analysts, risk managers, and operations staff: positions that once required degrees and offered stable career paths. Conor Hillery, JPMorgan Chase's co-CEO for Europe, offered a rare note of caution: "The one thing we have to be very careful about, in this rush and excitement about AI, is that people don't lose an understanding of the basics and fundamentals. Otherwise, we're storing up a big problem for the future." For anyone in finance, consulting, legal, or other knowledge work, the pattern is worth watching. The question isn't whether AI will reshape these industries, but how quickly the 30% efficiency gains banks are chasing become the expectation everywhere else.

QUICK TAKES

Baidu's AI Chip Unit Kunlunxin Files for Hong Kong IPO

The story: Baidu's AI chip subsidiary Kunlunxin has confidentially filed for a Hong Kong IPO as China pushes for semiconductor self-sufficiency. The company, which Baidu owns about 59% of, has shifted from being an internal supplier to selling chips to third-party customers like China Mobile. JPMorgan analysts forecast Kunlunxin's chip sales could increase sixfold to 8 billion yuan in 2026. Baidu's Hong Kong shares rose 7.5% on the news.

Your takeaway: US chip restrictions are forcing Chinese tech giants to accelerate homegrown alternatives. Kunlunxin's IPO shows how export controls are creating a new generation of domestic AI chip companies backed by major tech players.

DeepSeek Unveils Training Method That Could Shape Next-Gen AI Models

The story: Chinese AI startup DeepSeek published a research paper introducing "Manifold-Constrained Hyper-Connections," a new approach to training AI models that improves stability and reduces computational costs. CEO Liang Wenfeng co-authored the paper, which was uploaded directly to arXiv. Analysts at Counterpoint Research called it a "striking breakthrough" that could help bypass compute bottlenecks. The paper signals preparation for DeepSeek's next flagship model, expected around February.

Your takeaway: DeepSeek's research papers have historically preceded major model releases. This one suggests the company is finding ways to train powerful AI with fewer resources - the same approach that made their R1 model a surprise hit last year.

China Launches AI Science System Linked to 30+ Supercomputing Centers

The story: China launched an autonomous AI system on December 23 that can directly access the country's national supercomputing network and conduct scientific research independently. The platform connects over 30 supercomputing centers and serves more than 1,000 institutions. Officials say tasks that previously took researchers a full day can now be completed in about an hour. The system supports nearly 100 scientific workflows across materials science, biotechnology, and industrial AI.

Your takeaway: This launch comes weeks after Trump announced the US "Genesis Mission" for AI-led science. China's system is already operational and serving researchers, while the US plan faces a 270-day demonstration deadline.

RAND Study Examines How to Stop a Catastrophic Rogue AI

The story: A new RAND Corporation paper examines three options for responding to a globally distributed rogue AI: deploying a "hunter-killer" AI to destroy it, shutting down parts of the internet, or using nuclear-initiated EMP attacks to wipe out electronics. Each option carries significant risks and uncertain effectiveness. The study notes that current AI models already run across multiple data centers, making them difficult to shut down, and that advanced AI systems have shown self-preservation behaviors during testing.

Your takeaway: The paper's main conclusion is sobering: we have no good options if an advanced AI goes catastrophically wrong. The author argues governments are unprepared for worst-case scenarios and need to start planning now.

TOOLS ON OUR RADAR

🎤 BetterDictation Paid: Turn your voice into polished text with AI-powered dictation that's faster than typing.

🎯 Plus AI Paid: Add AI presentation powers directly inside Google Slides and PowerPoint, no new tool to learn.

📅 Motion Paid: Let AI auto-schedule your tasks into your calendar so you never have to plan your day again.

📧 Listmonk Free and Open Source: Send newsletters to millions of subscribers from your own server, powerful and completely free.

🥒 Pickle 1 Paid (Pre-order): AI smart glasses that remember your life, anticipate your needs, and think alongside you, all in a 68g frame with 12-hour battery. Note: This is a pre-purchase product not yet on the market and has not been tested.

TRENDING

Korean Team Develops AI for Personalized Cancer Vaccines - Researchers at KAIST created an AI model that identifies unique protein fragments in each patient's tumor to train their immune system against cancer recurrence. Clinical trials are planned for 2027.

Cybersecurity Skills Matter More Than Headcount - A survey of 16,000 security professionals found 88% experienced security incidents due to skills gaps, not staff shortages. AI and cloud security remain the most urgent skill needs for the second year running.

OpenAI Bets Big on Audio-First Devices - OpenAI is unifying teams to build audio-first personal devices expected in about a year that may resemble a pen. Former Apple design chief Jony Ive, who joined through a $6.5 billion acquisition, wants to reduce screen addiction.

China Surges Ahead in Humanoid Robotics - China's 150+ robotics companies are deploying affordable humanoid robots in factories and border patrols, while US firms face regulatory hurdles and slower commercialization timelines.

TRY THIS PROMPT (copy and paste into Claude, ChatGPT, or Gemini)

Podcast Episode Planner: Structure engaging episodes with segment timers, topic flow, guest prep, and show notes generation

What this does: Transforms chaotic podcast recording into structured production—with segment timing, topic flow planning, guest preparation checklists, and auto-generated show notes so your episodes feel professional instead of rambling.

What this looks like:

WHERE WE STAND(based on today’s Quick Takes and Trending news)

✅ AI Can Now: Train itself to identify cancer-specific protein fragments unique to each patient - something humans couldn't systematically catalog before.

❌ Still Can't: Be reliably shut down once distributed across multiple data centers - even with nuclear EMP attacks, success isn't guaranteed.

✅ AI Can Now: Complete a full day's worth of scientific research workflows in about an hour when connected to national supercomputing networks.

❌ Still Can't: Match Chinese humanoid robots on price - US models cost significantly more, limiting commercial deployment even when the technology works.

✅ AI Can Now: Train larger models with fewer resources by preventing the mid-training failures that waste weeks of compute time.

❌ Still Can't: Replace the cybersecurity skills gap - 88% of organizations experienced security incidents because they lack trained humans, not because they lack AI tools.

FROM THE WEB

RECOMMENDED LISTENING/READING/WATCHING

I'm Here (2010)

Spike Jonze directed this 30-minute short about two robots falling in love in Los Angeles. It's quiet and bittersweet, more interested in loneliness than technology. If you liked Her (same director), this feels like a sketch for those themes.

Thank you for reading. We’re all beginners in something. With that in mind, your questions and feedback are always welcome and I read every single email!

-James

By the way, this is the link if you liked the content and want to share with a friend.

Some links may be affiliate links. This helps support the newsletter at no extra cost to you.