GSV's AI News & Updates (03/17/25)
Manus Agent, Xbox Copilot, Apple AI, China AI Instruction, AI Nurse Pushback, Computer Operating API, AI Action Plan, Spain Regulates AI Content, "Yes-men on Servers"
General 🚀
Manus: An AI system designed for independent thinking, planning, and execution without human oversight, setting it apart from traditional AI assistants. Built on a multi-agent architecture, Manus operates like an executive overseeing specialized sub-agents, allowing it to handle complex workflows autonomously. It excels at analytical tasks with a defined scope, performing efficiently at a level comparable to a highly skilled human intern over a full workday.
Google Gemini just made two of its best features available for free: Deep Research and Gems are now available to all users, expanding access beyond the paid Gemini Advanced. Deep Research features agentic AI capability that conducts thorough, multi-step research on behalf of users.
Microsoft’s new Xbox Copilot will act as an AI gaming coach: Initially launching on the Xbox mobile app, Copilot for Gaming serves as a second-screen companion, offering real-time gameplay analysis, strategy tips, and in-game guidance. It can coach players (e.g., positioning in Overwatch 2), assist with game mechanics (e.g., crafting in Minecraft), and even locate materials or suggest cheats by seeing the player's screen.
Meta accelerates voice-powered AI push: The social media giant is introducing improved features as it bets on the technology driving the growth of so-called AI agents.
Google Gemini Can Now Use Your Search History to Provide Personalized Responses: Google says that search history will only be used when its reasoning models decide that it's helpful, and that early testers have found the feature useful for brainstorming and personalized recommendations.
Leaked Apple meeting shows how dire the Siri situation really is: The company is aiming to get some of the AI-powered features announced last June into iOS 19 — but it’s not guaranteed.
No part of Amazon is ‘unaffected’ by AI, says its head of AGI
China’s AI frenzy: DeepSeek is already everywhere — cars, phones, even hospitals
Education and the Future of Work 📚
China’s six-year-olds are already being offered AI classes in school in a bid to train the next generation of DeepSeek founders: Beijing schools will introduce AI courses for students as young as six, covering chatbot use, AI fundamentals, and ethics, with at least eight hours of AI instruction per year. The Ministry of Education has already piloted AI programs in 184 schools and announced plans for a 2025 white paper on AI education.
What should we do about teens using AI to do their homework? (podcast): A teenager explains why he shouldn’t have to write homework essays anymore. Is there some way for adults to force teens to still do homework? Or to convince them they should want to?
AI Helps Math Teachers Build Better "Scaffolds": The study introduced the first evaluation framework for lesson scaffolding based on expert teacher processes: observing student needs, developing instructional strategies, and implementing scaffolded lesson plans.
As AI nurses reshape hospital care, human nurses are pushing back: Hippocratic AI and others are developing AI nurses like "Ana" to handle admin tasks, answer patient queries, and monitor vitals 24/7, easing nurse workloads. However, National Nurses United warns AI is being used to "de-skill" and replace caregivers, compromising care quality.
How the AI Talent Race Is Reshaping the Tech Job Market: 36% of IT job listings in January were AI-related, particularly in tech giants, finance, and consulting firms. AI adoption is spreading beyond tech—retail, utilities, and healthcare are hiring for AI-integrated roles.
Startup nearly hired two fake AI-generated candidates posing as backend engineers using deepfake avatars in video interviews: These candidates passed technical interviews convincingly, raising suspicions only during non-technical final rounds when inconsistencies in their backgrounds and language skills emerged.
Sam Altman teases model with advanced creative writing capabilities: The model is said to demonstrate remarkable skill in metafiction, capturing nuanced themes of grief, memory, and AI consciousness with literary depth.
Sakana claims its AI-generated paper passed peer review — but it’s a bit more nuanced than that: The paper was peer-reviewed and accepted at an ICLR workshop before Sakana withdrew it, citing transparency and conference norms. Experts question its significance, arguing AI excels at mimicking human prose but lacks true scientific innovation.
AI Search Has A Citation Problem: Even when citing news, AI tools often pointed to syndicated versions rather than the primary publisher, depriving outlets of traffic and proper attribution.
Startups and Tools 🛠️
AI Startup Anysphere in Talks for Close to $10B Valuation: The maker of coding app Cursor surpassed $100M in recurring revenue in 12 months. Thrive Capital is expected to lead the funding round and the company has already raised $175M in previous venture rounds.
Cartesia Raises $64M to Advance Real-Time Voice: Their Sonic 2.0 model achieves ultra-low latency (90ms full models, 40ms real-time) and industry-leading voice cloning, capturing accents and tonal nuances.
ServiceNow buys AI agent company Moveworks for $2.85B: Moveworks’ AI assistant enhances ServiceNow’s automation by providing a front-end AI interface for task fulfillment and enterprise search. The acquisition positions ServiceNow against rivals like Salesforce, Microsoft, and Google in the growing AI agent and enterprise search market.
CoreWeave strikes $12B contract with OpenAI ahead of IPO, sources say
Tech 💻
OpenAI will let other apps deploy its computer-operating AI: The new Responses API comes with web search, the ability to look through files, and computer use out of the box
Alibaba Releases AI Model That Reads Emotions to Take On OpenAI: Alibaba’s Tongyi Lab has released R1-Omni, an open-source AI model capable of inferring human emotions from videos while also describing clothing and surroundings, enhancing computer vision capabilities.
Cohere’s Command A Pushes Enterprise AI Forward: A highly efficient and multilingual generative AI model designed for business applications. It supports 256,000 tokens (equivalent to 600 pages of text) and runs on just 2 GPUs (A100 or H100), unlike competitors requiring up to 32 GPUs.
Gemini 2.0 Flash native image generation: Generates illustrated stories while maintaining character and setting consistency. Supports multi-turn dialogue-based edits (conversational image editing), allowing iterative refinements.
How ‘chain of draft’ could cut AI costs by 90% while improving performance: CoD uses as little as 7.6% of the tokens required by Chain of Thought (CoT) reasoning, significantly cutting costs and computational load.
Snap introduces AI Video Lenses powered by its in-house generative model
Safety and Regulation ⚖️
AI Action Plan: Trump recently rescinded the Biden Administration's 2023 Executive Order on Artificial Intelligence (AI) and initiated efforts to develop a replacement "AI Action Plan."
American AI companies have very different ideas about regulation and how best to thwart China
Google calls for weakened copyright and export rules in AI policy proposal
OpenAI urges Trump administration to remove guardrails for the industry
Anthropic’s Recommendations to OSTP for the U.S. AI Action Plan
U.S. Likely to Ban Chinese App DeepSeek From Government Devices: Italy, South Korea, Australia, Canada, and Taiwan have restricted or banned DeepSeek due to privacy concerns.
Spain to impose massive fines for not labelling AI-generated content: Companies failing to label AI-generated content face fines of up to €35 million or 7% of global turnover. The law also bans subliminal AI tactics (e.g., chatbots promoting gambling, toys encouraging risks) and prohibits biometric classification, preventing AI from profiling individuals by traits or behavior.
Signal President Meredith Whittaker calls out agentic AI as having ‘profound’ security and privacy issues: AI agents require extensive, often unencrypted access to personal data, making them a major privacy and security concern. As companies push for AI-driven automation, users must weigh the risks of convenience vs. data vulnerability.
AI can steal your voice, and there's not much you can do about it: Five out of six leading AI voice cloning services have minimal safeguards, making it easy to impersonate someone without consent. AI voice cloning was used in the Democratic primaries for fake Biden robocalls, leading to a $6M fine and an FCC ban on AI-generated robocalls.
Americans are increasingly skeptical about AI's effects
Judge allows authors’ AI copyright lawsuit against Meta to move forward
Other
Hugging Face’s chief science officer worries AI is becoming ‘yes-men on servers’: AI today functions as "very obedient students" rather than scientific revolutionaries who challenge conventional thinking. AI isn’t designed to ask new, unconventional questions—it only memorizes and extrapolates existing reasoning patterns.
Nvidia’s next chips are named after Vera Rubin, astronomer who discovered dark matter
AI ‘wingmen’ bots to write profiles and flirt on dating apps