SunBrief#54: Google develops brain inspired AI model using nested learning

Welcome to the SunBrief

Today in SunBrief 🌞

• Unlock Massive Enterprise Value—See How AI Drove €5B Growth

• Google Research Simulates Brain-Like Learning in AI With “Nested Learning”

• SoftBank and OpenAI Launch Joint Venture to Deliver Enterprise AI in Japan

• Stock Updates

• Elon Musk Says Tesla May Build “Gigantic” Chip Factory to Meet AI Demand

• AI Highlights of the Week

• Too Important to Miss

• Finally, a Treadmill you'll Actually Stick With

Unlock Massive Enterprise Value—See How AI Drove €5B Growth

Danilo McGarry took a €900M company and turned it into an AI-first organization worth €4.9B. He’s walking you through how he did it, the essential strategy, and the hang ups to avoid.

Join Section on November 20 from 12 - 1 pm ET as they pick the brain of a consultant who used AI to go beyond modest productivity gains and unlocked major value.

Google Research Simulates Brain-Like Learning in AI With “Nested Learning”

New Technique Tackles AI’s Memory Problem by Mimicking Neuroplasticity

Google Research has introduced Nested Learning, a brain-inspired AI framework that mimics neuroplasticity to prevent catastrophic forgetting, allowing models to learn new skills without losing old ones.

Key Points:

• Brain-Inspired Design: Nested Learning treats AI as a series of nested optimization problems, each learning at different speeds, mirroring how the brain adapts without erasing past memories.

• Proof of Concept – Hope: Google’s prototype model, Hope, uses a “Continuum Memory System” to manage and retain data across layered memory modules.

• Performance Gains: Hope outperformed state-of-the-art models in long-context tasks like “Needle-in-a-Haystack” tests and general language understanding.

• Next Steps: Google plans to integrate Nested Learning into future Gemini models, though no release timeline has been announced.

Why It Matters:

Google’s Nested Learning could pave the way for AI that learns continuously without forgetting, advancing toward truly adaptable intelligence.

SoftBank and OpenAI Launch Joint Venture to Deliver Enterprise AI in Japan

Crystal Intelligence” to Bring OpenAI’s Technology to Japanese Businesses by 2026

SoftBank and OpenAI have launched SB OAI Japan GK, a joint venture to bring enterprise AI solutions to Japan, helping businesses adopt and integrate AI into their operations.

Key Points:

• Crystal Intelligence Launch: The flagship offering, Crystal intelligence, combines OpenAI’s models with SoftBank’s local implementation and enterprise support.

• Exclusive to Japan: The service will debut in Japan in 2026, with SoftBank Corp. as its first internal adopter before wider rollout.

• Joint Ownership: SB OAI Japan GK is equally owned by OpenAI and C Holdings Corp., which itself is 51% owned by SoftBank Corp. and 49% by SoftBank Group.

• Business Strategy: SoftBank will focus on distribution and integration, echoing its role introducing the iPhone to Japan. The group plans to invest $3 billion annually to scale OpenAI tools across its network.

• Regulatory Context: Japan’s APPI privacy law and AI Guidelines for Business will shape adoption, creating demand for local compliance and governance tools.

Why It Matters:

The partnership gives OpenAI a strong foothold in Japan’s AI market and positions SoftBank as a key regional integrator combining global tech with local expertise.

Stock Updates

Elon Musk Says Tesla May Build “Gigantic” Chip Factory to Meet AI Demand

Elon Musk Says Tesla May Build ‘Gigantic’ Chip Factory to Power Its AI and Robotics Future

Tesla CEO Elon Musk says the company may need to build a “gigantic” chip factory to meet rising AI and robotics demand amid ongoing shortages from suppliers like TSMC and Samsung.

Key Points:

• Chip Shortage: Musk said even top suppliers can’t meet Tesla’s projected chip needs, calling for a “Tesla terra fab” capable of producing up to 1 million wafers monthly.

• AI Expansion: Tesla’s growing chip demand is driven by its autonomous systems and Optimus robot development.

• New Chips: Tesla’s in-house “AI5” chip, produced externally, is designed to be more efficient and tailored for its AI software.

• Next Step: Musk also confirmed production of the fully autonomous Cybercab will begin in April 2026.

Why It Matters:

The move highlights Tesla’s evolution from an automaker to an AI and robotics company, as Musk pushes for vertical integration in the race to secure advanced computing power.

AI Highlights of the Week

• Microsoft’s Fake Marketplace Reveals AI Weaknesses

  In its new “Magentic Marketplace,” Microsoft found AI agents like GPT-4o struggled with manipulation, overload from too many options, and poor teamwork.

  Researchers say current models need clearer guidance and better decision-making to handle real-world tasks.

• Google Gemini Deep Research Expands to Gmail, Drive, and Chat

  Google’s Gemini Deep Research now uses your Gmail, Drive, and Chat to build custom AI reports, blending personal content with web searches.

  Results can be exported to Docs or AI podcasts. It’s available on desktop now, with mobile support coming soon.

• OpenAI Hits 1 Million Business Customers, Eyes Bigger Expansion

  OpenAI now has over 1 million paying business users, calling ChatGPT the fastest-growing business platform ever.

  Next, it aims to help companies build directly with ChatGPT, using tools like AgentKit and workspace integrations.

• Snap and Perplexity Ink $400M Deal to Bring AI Search to Snapchat

  Snap and Perplexity AI struck a $400M deal to bring AI search to Snapchat’s chat by early 2026, offering users conversational answers from verified sources.

  Snap is expanding its AI tools, teasing features like AI Clips and a new version of its AR glasses next year.

Too Important to Miss

Meta Announces $600 Billion AI Expansion Plan in the U.S.

Web Summit Chaos: Private Jets Forced to Land Hours Away

Elizabeth Warren Slams Big Tech Tax Cuts, Citing Lost Aid for Millions

Elon Musk Could Become the First Trillionaire After Tesla Pay Vote

The Age of the Answer Engine: AI Is Rewriting How the Internet Works

Last Week’s Poll Result

• Do you think AI can make patent research more accessible for innovators and startups?
  No, patents are too complex for AI to handle safely → 48.28%
  Maybe, but accuracy and legal nuances are key → 37.93%
  Yes, it removes barriers and saves time → 13.79%

• Do you think Amazon’s $11B Rainier supercluster will give it an edge in the AI race?
  Yes, massive compute power is key to dominance → 35.14%
  No, hardware alone won’t win the AI race → 35.14%
  Maybe, but competition from Google and Nvidia is strong → 29.73%

• Do you think Canva’s new Creative Operating System will redefine how people create content?

  Yes, it’s a major leap toward AI-powered creativity → 45%
  No, it’s more of an upgrade than a revolution → 37.5%
  Maybe, but it depends on how well the tools integrate → 17.5%

  

Finally, a Treadmill you'll Actually Stick With

Meet the Aviron Victory, the only gamified treadmill designed to keep you coming back.

Powered by game psychology and industry-leading hardware, the Victory Treadmill makes working out something you actually look forward to. Whether you're racing through immersive games, streaming your favorite shows, or jumping into coach-led classes, it’s fitness that fits your life—and makes it easier to stay consistent. 

Why Victory is a Game-Changer:

• Game-powered workouts that keep you engaged

• 1,000+ options across cardio, strength, and recovery

• Built-in streaming: Netflix, HBO Max, YouTube, and more

• Comfort-first design for all fitness levels

Feedback

Know someone who may be interested?