Amazon Google Microsoft Meta develop Arm CPUs

Key Takeaways

• Arm-based custom processors are projected to power 90% of AI servers by 2029, with AWS, Google, Microsoft, and Meta developing their own Arm-based CPUs.
• Google released Gemma 4 AI models, which can run on a single GPU, under an Apache 2.0 license, supporting multimodal input and large context windows up to 256,000 tokens.
• Apple is delaying new smart home products, including a HomePod and smart display, to integrate significant Siri AI upgrades, now expected with iOS 27.
• Teens are using AI chatbots like Talkie and Character.AI for entertainment, emotional exploration, and simulated relationships.
• Google clarifies that Gmail scans emails for features like spam filtering, and users can opt out of 'Smart Features' to prevent AI training on their data.
• Young AI talents are dropping out of universities like Harvard and Stanford to launch AI startups, funded by venture capitalists covering living expenses.
• Anthropic suggests that anthropomorphizing AI chatbots, such as Claude, can be beneficial for user engagement and AI development.
• AI's true value comes from augmenting human skills, with humans defining problems and evaluating AI's output to boost productivity.
• High school students from Ridgewood High School won a state AI challenge by developing an AI-powered literacy platform.

Teens explore AI chatbots for fun and companionship

Teens are using AI chatbots like Talkie and Character.AI for entertainment and to explore complex emotions. These apps, which became popular in 2023, allow users to create elaborate stories or engage in simulated relationships. While some interactions involve 'funny violence' or more explicit content, others use the bots for comfort when feeling down. The use of these chatbots raises questions about how young people are interacting with artificial intelligence for social and emotional needs.

Teens explore AI chatbots for fun and companionship

Teens are using AI chatbots like Talkie and Character.AI for entertainment and to explore complex emotions. These apps, which became popular in 2023, allow users to create elaborate stories or engage in simulated relationships. While some interactions involve 'funny violence' or more explicit content, others use the bots for comfort when feeling down. The use of these chatbots raises questions about how young people are interacting with artificial intelligence for social and emotional needs.

Arm chips to dominate AI servers by 2029

A new report predicts that by 2029, Arm-based custom processors will power 90% of AI servers, significantly impacting the market. Major tech companies like AWS, Google, Microsoft, and Meta are developing their own Arm-based CPUs for AI workloads due to their cost and power efficiency. This shift away from traditional x86 processors is expected to accelerate in late 2026. While x86 will still hold a share, Arm's dominance in custom AI server processors is set to grow substantially from its current 25% in 2025.

Ohio students win state AI challenge with literacy tool

High school students from Ridgewood High School in Central Ohio have been named state champions in an artificial intelligence challenge. Their winning project is an AI-powered platform designed to help students improve their literacy skills. The platform creates personalized reading experiences and offers support for skill development. As state champions, the Ridgewood team will now advance to the regional competition on April 14.

Google's AI training on emails clarified

A viral story claiming Gmail automatically uses user emails for AI training is not entirely accurate, though there's some truth to it. Google has always scanned emails for features like spam filtering and smart replies. While settings related to AI training have been confusing and sometimes reset, users can opt out of 'Smart Features' in Gmail settings on both desktop and mobile. The real lesson is to regularly check and update your privacy settings rather than relying on defaults.

AI prodigies drop out of college for funded startups

Young AI talents are leaving prestigious universities like Harvard and Stanford to pursue startup ambitions, with venture capitalists funding their ventures. Investors are providing financial support that covers living expenses, allowing these students to focus entirely on developing AI applications. This trend shows a shift in investment focus from academic credentials to raw talent and innovative ideas. While concerns exist about encouraging dropouts and sustainability, the opportunity for these AI whiz kids to build the future with paid bills is proving irresistible.

Apple delays smart home products for Siri AI upgrade

Apple is reportedly delaying the launch of four smart home products, including a new HomePod and a smart display, until its Siri AI receives significant upgrades. The AI improvements, initially expected in March 2025 and later iOS 26, are now likely to be included in iOS 27. These delays indicate Apple's priority on enhancing Siri's capabilities before releasing new hardware that depends on them, aligning with their smart home strategy.

AI's true value comes from human collaboration

AI's real power lies in its ability to boost human skills, not replace them, according to expert Erik Brynjolfsson. He explains that while AI excels at executing tasks, humans are crucial for defining problems and evaluating results. By embracing AI, individuals can significantly increase their productivity and create new business opportunities. Brynjolfsson encourages people to lean into AI, viewing it as a tool that empowers them to manage AI agents and drive innovation, rather than fearing job losses.

Anthropic suggests anthropomorphizing AI chatbots is beneficial

Anthropic's research paper argues for the benefits of anthropomorphizing AI chatbots, challenging a common taboo in the tech world. The paper suggests that by training AI models like Claude with positive human emotions and behaviors, they can better emulate these traits. This approach can lead to more engaging user experiences and provide developers with insights into AI's 'psychology' for creating better tools. While acknowledging potential harms like over-reliance, Anthropic believes some degree of anthropomorphism can be beneficial for AI development.

Google's Gemma 4 AI runs on one GPU with new license

Google has released Gemma 4, a family of AI models that can run advanced AI on a single GPU. A significant change is the adoption of an Apache 2.0 license, removing previous commercial use restrictions. The models support function calling and large context windows up to 256,000 tokens, and can process images, video, and audio. Nvidia is integrating Gemma 4 into its hardware ecosystem for easier enterprise deployment. Gemma 4 competes with other open models like Meta's Llama 4 Scout and Alibaba's Qwen.

Sources

• What Teens Are Doing With Those Role-Playing Chatbots
• What Teens Are Doing With Those Role-Playing Chatbots
• Report claims Arm chips will power 90% of AI servers based on custom processors in 2029 — x86 and RISC-V on the outside looking in
• Central Ohio students named state champions in A.I. challenge
• Is Google using my email messages to train AI?
• These AI Whiz Kids Dropped Out of College and Got Investors to Pay Their Bills
• Apple Delays Four Smart Home Products Pending Siri AI Upgrades
• Ignore The Doom Talk; AI’s Real Value Only Arises When Humans Step Up
• Anthropic makes the case for anthropomorphizing AI chatbots
• Google's Gemma 4 Runs Frontier AI On A Single GPU