google, apple and meta Updates

Google Research recently unveiled a significant advancement in artificial intelligence called Nested Learning on November 7, 2025. This innovative method aims to solve the persistent problem of 'catastrophic forgetting' in AI models, allowing systems like Google's proof-of-concept 'Hope' to learn continuously without losing previously acquired knowledge. The approach mimics human memory by treating AI as many interconnected learning processes that update at varying speeds, a crucial step towards achieving artificial general intelligence (AGI). Google's commitment to AI extends beyond research, as the company raised its 2025 capital budget to $92 billion and expects to spend around $100 billion in 2026, also hosting a conference in New York City to discuss the complex topic of AI consciousness. Meanwhile, other tech giants are making massive strides and investments. Apple CEO Tim Cook announced on November 9, 2025, that the next version of Siri will soon launch with enhanced AI capabilities, making it smarter, more personalized, and able to operate iPhone apps. Meta Platforms Inc. is planning an enormous $600 billion investment in U.S. artificial intelligence infrastructure over the next decade, aiming to significantly boost its AI computing power by acquiring advanced AI chips, including many graphics processing units (GPUs) from Nvidia, and developing new data centers. Amazon is also expanding its footprint, with plans to double its data center capacity by 2027. The broader AI landscape shows a surge in spending, with Bank of America predicting global hyperscale spending to increase by 67% in 2025 and 31% in 2026, reaching $611 billion. Experts forecast annual AI data center investment could triple to over $1.2 trillion by 2030. This intense competition is also reflected in the job market, where demand for AI skills is rapidly growing, with job postings mentioning AI skills increasing by 16% in just three months. Experts are also emphasizing the importance of knowledge elicitation techniques to embed deep, unwritten human expertise into generative AI models, ensuring highly skilled systems. Globally, the competition is heating up. Open-source Chinese AI models, such as Kimi K2 from Beijing-based Moonshot AI, are gaining popularity in Silicon Valley due to their lower cost and strong performance, attracting companies like Airbnb and Cognition AI. However, the United States faces a challenge in providing the necessary energy for this AI expansion, lagging behind China, which added 429 GW of new power capacity in 2024—six times more than the U.S., focusing on renewables. In education, the Utah State Board of Education's Standards and Assessment Committee approved new Career and Technical Education course standards that include a separate section on the 'responsible use' of artificial intelligence and require students to 'verify AI sources' to avoid misleading results.

Key Takeaways

• Google Research introduced Nested Learning and the HOPE AI model on November 7, 2025, to enable continuous AI learning without forgetting old information, aiming for artificial general intelligence (AGI).
• Apple CEO Tim Cook announced on November 9, 2025, that the next version of Siri will soon launch with AI capabilities for enhanced personalization and iPhone app operation.
• Meta Platforms Inc. plans to invest $600 billion in U.S. AI infrastructure over the next decade, including purchasing Nvidia GPUs, to boost its AI computing power.
• Global hyperscale AI spending is projected to increase by 67% in 2025 and 31% in 2026, reaching $611 billion, with Google raising its 2025 capital budget to $92 billion.
• Amazon's data center capacity is set to double by 2027, reflecting the significant expansion in AI infrastructure.
• Demand for AI skills in the job market is rapidly growing, with job postings mentioning AI skills increasing by 16% in three months.
• Open-source Chinese AI models are gaining popularity in Silicon Valley due to their lower cost and strong performance, attracting U.S. companies.
• The U.S. is falling behind China in providing the energy needed for the growing AI industry, with China adding six times more new power capacity in 2024, primarily from renewables.
• Experts are utilizing knowledge elicitation techniques to embed deep, unwritten human expertise and best practices into generative AI and large language models.
• Utah's State Board of Education approved new CTE course standards, including a section on the responsible use of AI and requiring students to verify AI sources.

Google unveils Nested Learning to fix AI memory loss

Google Research introduced a new method called Nested Learning on November 7, 2025. This approach helps AI models, like the proof-of-concept "Hope," learn continuously without forgetting old information. Catastrophic forgetting is a major problem where AI erases past knowledge when learning new things. Nested Learning works by treating AI as many interconnected learning processes that update at different speeds, similar to how human memory functions. This aims to create more dynamic and efficient AI systems that can improve over time without needing to be retrained from scratch.

Google's HOPE AI learns nonstop with Nested Learning

Google researchers developed a new machine learning model called HOPE, which uses a novel concept called Nested Learning. This model aims to help AI learn continuously without forgetting past information, a major challenge for large language models (LLMs). HOPE has a self-modifying design that improves long-context memory management. Google believes Nested Learning can help achieve artificial general intelligence (AGI) by allowing AI to integrate new experiences like the human brain. The research findings were published at NeurIPS 2025 in a paper titled "Nested Learning: The Illusion of Deep Learning Architectures."

Google AI learns continuously like the human brain

Google created a new artificial intelligence system that learns continuously, much like the human brain. This breakthrough uses a method called Nested Learning. The system can gain new knowledge and skills over time without overwriting or losing information it learned before. This allows the AI to integrate new information smoothly, making it more adaptable and robust. This continuous learning capability is a significant step towards creating more effective AI systems for fields like robotics, personalized education, and advanced data analysis.

Utah approves new CTE course standards with AI rules

The Utah State Board of Education's Standards and Assessment Committee approved new Career and Technical Education (CTE) course standards. These standards cover 39 courses and will now go to the full board for consideration. Committee members proposed adding a separate section on the "responsible use" of artificial intelligence to any course where AI is recommended. They also made changes to business communications courses, requiring students to "check sources for credible and accurate information" and "verify AI sources" to avoid misleading results. Edits were also approved for Construction Trades and introduction-to-drones materials.

Experts share hidden knowledge with AI models

Dr. Lance B. Eliot explains how to use knowledge elicitation techniques to add deep expertise and best practices into generative AI and large language models (LLMs). This method helps uncover hidden rules of thumb from human experts that are often not written down. While some AI experts might think these techniques are old, they are crucial for making LLMs highly skilled in specific areas like medicine or law. This approach goes beyond just feeding documents to AI, ensuring that the "secret sauce" of expert knowledge is included to create truly expert AI systems.

Apple CEO says new Siri will use AI soon

Apple CEO Tim Cook announced that the next version of Siri will launch soon. He revealed this during an investor call on November 9, 2025. The new Siri will use artificial intelligence to become smarter and more personalized. It will also be able to operate iPhone apps using this new AI technology. This update aims to enhance the voice assistant's capabilities significantly.

Meta plans 600 billion dollar AI investment in US

Meta Platforms Inc. plans to invest a massive $600 billion in U.S. artificial intelligence infrastructure over the next decade. This significant investment will boost Meta's AI computing power by acquiring advanced AI chips and developing data centers. The company aims to compete strongly with other tech giants like Google, Microsoft, and Amazon. Meta will purchase many graphics processing units (GPUs) from Nvidia and might also develop its own custom AI chips. This move signals Meta's long-term vision for AI leadership and its belief in the technology's transformative potential.

AI spending surges as tech giants invest trillions

Investing.com reports that spending on artificial intelligence infrastructure continues to rise sharply. Bank of America now predicts global hyperscale spending will increase by 67% in 2025 and 31% in 2026, reaching $611 billion. Google alone raised its 2025 capital budget to $92 billion and expects to spend about $100 billion in 2026. Amazon's data center capacity is set to double by 2027. Experts believe annual AI data center investment will triple to over $1.2 trillion by 2030, showing a strong commitment from tech companies to scale AI compute power in a race that resembles an arms buildup.

AI skills are key for jobs in a tough market

While overall tech hiring is down, demand for artificial intelligence skills is rapidly growing. Job postings mentioning AI skills increased by 16% in three months, according to ManpowerGroup. Companies seek employees who can interpret AI output, spot bad data, and integrate AI insights into business decisions. Salaries for machine learning engineers have risen sharply, much more than for software engineers. However, human skills like curiosity, fast learning, and strategic thinking remain important, especially when combined with AI tool knowledge, as the AI landscape evolves quickly.

Google hosts conference on AI consciousness

Google recently held a conference in New York City to discuss the possibility of artificial intelligence consciousness. Scientists and researchers, including Jonathan Birch from the London School of Economics and Jeff Sabo from New York University, attended the event. This marks a shift for Google, which previously dismissed such discussions. Other AI companies like Anthropic are also seriously studying AI consciousness. The conference explored how to make AI safe and beneficial for all, including animals, showing a growing focus on these important questions.

Chinese AI models gain ground in Silicon Valley

Open-source Chinese AI models are becoming popular in Silicon Valley due to their lower cost and strong performance. Venture capitalist Chamath Palihapitiya noted that his company uses Kimi K2 from Beijing-based Moonshot AI because it is much cheaper than US alternatives. Other US companies like Airbnb and Cognition AI are also reportedly using or exploring Chinese AI. While geopolitical concerns exist, developers are attracted to the affordability and ability to fine-tune these models locally. The US still leads in advanced chips, but China's open-source strategy is attracting many developers, prompting questions about why Silicon Valley is switching sides.

US lags China in AI energy and renewables

The United States is falling behind China in providing the energy needed for the growing artificial intelligence industry. China added 429 GW of new power capacity in 2024, six times more than the US, focusing on solar, wind, nuclear, and gas. The US, however, relies on aging and expensive coal plants. Experts warn that without more energy, the US risks becoming a consumer rather than an innovator in AI. Data centers could help by being more flexible with their electricity use, but a larger shift to renewable energy is crucial. If China continues its rapid green energy buildout, it could leapfrog the US in AI progress.

Sources

• AI Memory: Google Research Unveils Nested Learning, a New AI Paradigm to Overcome 'Catastrophic Forgetting'
• Google touts big step toward continual learning with new AI model ‘HOPE’
• Google mimics brain flexibility in a new AI that learns nonstop
• Standards panel approves CTE course standards with new AI guidance and multiple course edits
• Using Knowledge Elicitation Techniques To Infuse Deep Expertise And Best Practices Into Generative AI
• Apple CEO reveals: New Siri will include artificial intelligence
• Meta to Invest $600 Billion in U.S. AI
• AI’s capital expenditure shows no sign of cooling By Investing.com
• The dimming job market's bright spot: AI skills
• Google Holds Conference On AI Consciousness With Scientists And Researchers
• How Much of Silicon Valley Is Built on Chinese AI?
• The State of AI: Energy is king, and the US is falling behind