GSV's AI News & Updates (11/25/24)

Apple’s LLM Siri, Mistral’s Pixtral Large, Google Gemini Memory, GPT-4 Beats Doctors, Meta’s Llama Pivot, Microsoft AI Agents, Teams Language Interpreter, X sues Deepfake Deception Act

Join me at the leading edge of all things AI in EDU at the second annual AI SHOW @ ASU+GSV, where 12,000+ PreK to Gray attendees will experience cutting-edge AI innovations through live EdTech demos, promptathons and hackathons, hands-on workshops, incredible thought leadership, and so much more — April 5-7 at the San Diego Convention Center. Register for FREE now, capacity is limited.

---

General 🚀

• Apple is reportedly working on ‘LLM Siri’ to compete with ChatGPT”: Expected to debut with iOS 19 and macOS 16 at WWDC 2025, LLM Siri will feature enhanced conversational abilities, faster processing of complex requests, better text generation, and improved integration with third-party apps.

• Mistral unveils new AI models and chat features:

  • Le Chat Chatbot Enhancements: Matches ChatGPT (on paper) on web search, image generation, and an interactive canvas tool for content editing, multimodal document processing including PDFs and images, and Black Forest Labs' Flux Pro image generation.

  • Two advanced AI models: Pixtral Large, a 124-billion-parameter multimodal powerhouse that outperforms competitors in processing text and images, and Mistral Large 24.11, an updated text model with enhanced long-context understanding and document analysis capabilities

    

• A Chinese lab has released a ‘reasoning’ AI model to rival OpenAI’s o1: DeepSeek-R1-Lite-Preview outperforms OpenAI's o1 model on benchmarks like AIME and MATH, employing chain-of-thought reasoning to tackle complex problems while providing transparent step-by-step thought processes. The model adheres to strict censorship protocols common in Chinese AI technology, avoiding sensitive political topics.

• Google’s Gemini chatbot now has memory: Google’s Gemini chatbot now includes a memory feature that can remember users' preferences, life details, and work-related info to provide personalized responses, similar to ChatGPT’s memory feature.

• Microsoft Teams' AI Language Interpreter Will Mimic Your Voice: Microsoft is testing an AI interpreter that can simulate speakers' voices in near-real-time across nine languages for its Teams platform. This feature is expected to launch broadly in 2025 for users with a Microsoft 365 Copilot license. The AI interpreter aims to democratize access to high-quality translations, reducing reliance on expensive human interpreters.

• A.I. Chatbots Defeated Doctors at Diagnosing Illness: In a new study, GPT-4 outperformed doctors in diagnosing complex medical cases. The AI scored an average of 90%, beating both doctors alone (74%) and doctors who used AI (76%).

• How Mark Zuckerberg has fully rebuilt Meta around Llama: Llama powers a wide range of Meta products, including AI assistants, advertiser tools, and augmented reality features in Ray-Ban glasses and Quest headsets. Meta's multibillion-dollar investment in AI integration across its platforms has helped the company recover from its metaverse challenges and restore industry credibility.

• Menlo Ventures 2024: The State of Generative AI in the Enterprise: Spending on generative AI applications reached $4.6B in 2024 (8x growth). Investments totaled $6.5B, but a multi-model approach is becoming standard. OpenAI’s market share declined to 34%, with Anthropic gaining traction (24%). Enterprises are increasingly developing in-house tools (47% vs. 20% in 2023), reflecting growing confidence in internal capabilities.

• Business spending on AI surged 500% this year to $13.8 billion:

  • OpenAI's enterprise market share dropped from 50% to 34%, while Anthropic doubled its share to 24%, driven by advancements like Claude 3.5. Companies increasingly use multiple AI models, selecting the best fit for specific use cases.

  • AI agents, capable of performing multistep tasks and automating workflows, emerged as a leading trend and investment focus in 2024.

• Wall Street's Elites Are Piling Into a Massive AI Gamble: The demand for digital infrastructure to support AI, including data centers and energy grids, is fueling massive investments, estimated at $1-2 trillion. Tech giants like Google, Microsoft, and Amazon (known as "hyperscalers") are driving demand, spending over $52.9 billion on AI infrastructure in just three months.

Education and the Future of Work 📚

• Start-Up Founder Who Sold A.I. Chatbot to Schools Is Charged With Fraud: Joanna Smith-Griffin, founder of AI education start-up AllHere Education, was charged with wire fraud, securities fraud, and aggravated identity theft for misrepresenting revenue and customer data to investors, raising nearly $10M fraudulently.

• Microsoft Signs AI-Learning Deal With News Corp.’s HarperCollins: They will use select nonfiction backlist titles for training AI models, focusing on improving model quality and performance. The deal includes limited scope and guardrails to respect authors' rights, allowing authors to opt in or out. The content will not be used to generate new books without human authors.

• ‘PDF to Brainrot’ study tools are a strange iteration on a TikTok trend: AI study tools convert uploaded text into audio paired with visually engaging, “oddly satisfying” videos (e.g., Minecraft gameplay, ASMR clips). Popular on TikTok, these tools cater to students who prefer multitasking by watching these videos while listening to their study material.

• Massachusetts student's punishment for AI use can stand, US judge rules: A federal court ruled against parents who sued a Massachusetts school district for disciplining their son for using AI improperly on an assignment. The student copied and pasted text generated by an AI tool, Grammarly, into a project without citation. Tools like Turnitin flagged portions of the project as AI-generated.

• Microsoft quietly assembles the largest AI agent ecosystem—and no one else is close: Microsoft has rapidly built the largest enterprise AI agent ecosystem, with over 100,000 organizations using Copilot Studio to create or edit AI agents. Microsoft's AI agents automate business tasks, support workflows, and integrate with existing IT landscapes.

  • Microsoft reveals AI employees at Ignite — agents will come to the workplace

• Growing Up: Navigating Generative AI’s Early Years – AI Adoption Report: Gen AI adoption has surged, with weekly usage increasing from 37% in 2023 to 72% in 2024. It is widely used across functions like document editing, data analytics, and summarization. However, its impact is highest in IT (58% see it as highly impactful). While 72% of companies plan to increase budgets next year, growth is expected to slow as companies seek clear ROI.

  

Tech 💻 

• AI’s Slowdown Is Everyone Else’s Opportunity: The once-reliable "scaling laws," which suggested that more data and computing power would exponentially improve AI models, are showing signs of diminishing returns. As the pace of capability growth slows, AI developers are exploring new methods, like improving models during the inference phase (e.g., OpenAI’s o1), rather than relying solely on more data and computing power. This could shift the focus toward reasoning, safety, and fairness in AI.

• Nvidia’s CEO defends his moat as AI labs change how they improve their AI models: Huang and other industry leaders, like Anthropic CEO Dario Amodei, stated that AI pretraining scaling remains intact, though diminishing returns are emerging. Huang noted the eventual growth of AI inference as more users adopt AI models.

• The Bitter Religion: AI’s Holy War Over Scaling Laws: The AI community is locked in a doctrinal battle about its future and whether sufficient scale will create God.

• Between the Booms: AI in Winter: After people stopped caring, artificial intelligence got more interesting. “The evolution of AI mirrors the S-curve of technological change: slow beginnings, rapid growth, and eventual transformation into broader applications."

• Are the robots finally coming?: Advances in physical AI mean machines are learning skills previously thought impossible.

Safety and Regulation ⚖️

• US Justice Department Seeks to Unwind Google’s Anthropic Deal: The proposal aims to bar Google from acquiring, investing in, or collaborating with companies influencing consumer search, including query-based AI products. It also calls for divesting Chrome, which a judge ruled reinforced Google’s dominance.

  • OpenAI Considers Taking on Google With Browser

• US government commission pushes Manhattan Project-style AI initiative: The USCC recommends a Manhattan Project-style initiative to develop artificial general intelligence (AGI) that matches or surpasses human intelligence, emphasizing public-private partnerships but offering no specific investment strategies. OpenAI, which recently released a U.S. AI strategy blueprint, has also advocated for increased government funding to accelerate AI advancements.

• Musk’s X Sues to Block California’s Deepfake Deception Act: The Defending Democracy From Deepfake Deception Act of 2024 (AB 2655) requires online platforms to remove AI-generated deceptive election content. The complaint claims the law could lead to censorship of political speech, emphasizing that the First Amendment protects even potentially false criticisms of government officials and candidates.

Other

• Pokémon Go Players Have Unwittingly Trained AI to Navigate the World: Niantic is developing a "Large Geospatial Model" (LGM) using data collected from millions of players of its AR games, like Pokémon Go, to enable computers to perceive, navigate, and interact with physical spaces.

• Coca-Cola causes controversy with AI-made ad: The advertisement, which is a remake of Coca-Cola's classic 1995 holiday commercial "Holidays Are Coming," was created using generative AI technology. Despite initial positive reactions, the ad quickly faced criticism for being perceived as "soulless" and lacking emotional depth.

  

• David Attenborough “Profoundly Disturbed” By AI Clone Of His Voice