GSV's AI News & Updates (06/26/24)
Claude Sonnet 3.5, Speak $20M Series B-3, Emergence.ai, Microsoft and Google EDU Releases, Amira and Istation, NoRedInk AI Grading, DeepMind V2A, Runway Gen-3 Alpha
General 🚀
Anthropic Releases Claude Sonnet 3.5: It’s the first in their new Claude 3.5 model family and represents a significant advancement in AI capabilities, outperforming competitor models (GPT-4o and Gemini 1.5 Pro) and Anthropic's previous Claude 3 Opus on various benchmarks. They also dropped Artifacts, which allows users to see, edit, and build upon Claude's creations (diagrams, infographics, games) in real-time, enhancing its utility as a collaborative work environment. Anthropic recently hired Instagram’s co-founder Mike Krieger as their CPO.
AI’s $600B Question: The AI bubble is reaching a tipping point. Navigating what comes next will be essential.
Roblox’s Road to 4D Generative AI: They aim to create interactive digital content that goes beyond traditional 3D modeling by incorporating functionality, interactivity, and realistic behavior within virtual environments.
Perplexity AI Results Include Plagiarism and Made-Up Content, Reports Say: Wired conducted an investigation revealing that Perplexity likely accessed their site and other Condé Nast properties thousands of times without permission, despite being blocked by the Robots Exclusion Protocol.
OpenAI's GPT-5 Pushed Back To Late 2025, But Promises Ph.D.-Level Abilities
Education and the Future of Work 📚
Language Learning App Speak Raises $20M in Series B-3 Financing and Hits $500M Valuation: Speak differentiates itself from competitors by focusing on spoken language practice and repetition rather than traditional vocabulary and grammar memorization. The app now boasts over 10M users across more than 40 countries, with its user base doubling annually for the past five years.
Satya Nitta, Former CEO of Merlyn Mind, co-founds Emergence.ai: They have secured $97.2M in investment from Learn Capital and $100M+ in credit lines. Emergence aims to revolutionize AI agent technology by developing an 'agent-based' system that utilizes generative AI models to automate tasks typically performed by knowledge workers.
Microsoft EDU Features and Improvements:
Guided Content Creation: AI to help educators get started with creating materials like assignments, lesson plans, lecture slides, and more
Quiz generation through Copilot in Forms: Educators can tell Copilot what they want to test, such as specific topics covered in class, and Copilot will create a well-crafted quiz draft with correct answers.
Suggested AI Feedback for Educators: Educators can review, edit, or discard suggested student feedback, which is based on individual student progress, rubrics selections, or assignment instructions.
Flashcards generation: the creator can review the flashcards created, add hints, add images, and then practice or share with their students or peers. Practicing flashcards always ends with a summary of what the learner already knows and what may need additional repetition.
Speaker Progress: Speaker Progress helps students develop confidence in their presentation skills and reduces anxiety by providing real-time coaching and feedback on their public speaking skills like their pace, pitch, filler words, and more.
New AI tools for Google Workspace for Education:
Gemini for Google Workspace: This AI assistant helps educators save time, create engaging learning experiences, and inspire new ideas.
Google Vids: A collaborative video creation app that makes it easy for educators and students to create engaging videos.
Read Along in Google Classroom: This AI-powered tool helps students build independent reading skills with real-time support.
Gemini in Classroom (pilot): for easier AI lesson and content planning.
AI training for educators: Google offers a two-hour, self-paced course called "Generative AI for Educators" to help teachers learn how to use AI tools effectively in their work.
Google brings its Gemini AI to Gmail to help you write and summarize emails
Amira Learning And Istation Merge, Expanding K-12 AI Solutions: Amira's sophisticated speech recognition and AI-driven tutoring software will complement Istation's assessment platform, content library, and distribution network. The merged company will serve 4M K-12 students across all 50 U.S. states and 17 countries.
NoRedInk Expands Its AI-Powered Grading and Feedback Functionality to All Customers: NoRedInk, a leading provider of online writing curriculum used in over 60% of U.S. school districts, has announced the expansion of its AI-powered Grading Assistant tool to all NoRedInk Premium teachers. The Grading Assistant utilizes AI to evaluate student writing, providing timely and consistent feedback.
Coursera Launches AI Plagiarism Detector: Coursera is introducing several other AI-powered features to enhance academic integrity: AI-assisted grading, proctoring and lockdown browser, AI-based Viva Exam, quiz prep and AI Coursera Coach.
LinkedIn Learning adds more AI-powered coaching features: The new pilot program will offer users AI-delivered advice, crafted based on wisdom from particular LinkedIn Learning instructors. The AI feature will enable users to get instantaneous responses—and may save instructors from repeatedly fielding common questions that many users have.
GPTs are GPTs: An Early Look at the Labor Market Impact Potential of Large Language Models (Research)
A Generation of AI Guinea Pigs: Children are once again serving as beta testers for a new generation of digital tech, just as they did in the early days of social media. Different age groups will experience AI in different ways—the smallest children may hear bedtime stories generated via ChatGPT by their parents, while older teens may run into chatbots on the apps they use every day—but this is now the reality.
A New Digital Divide: Student AI Use Surges, Leaving Faculty Behind: While both students and faculty have concerns with generative artificial intelligence, two new reports show a divergence in AI adoption.
Homeschoolers Embrace AI, Even As Many Educators Keep It at Arms’ Length
Startups and Tools 🛠️
HeyGen Raises $60M Series A: The funding round values the company at over $500M. HeyGen has experienced remarkable growth, increasing its ARR from $1M to over $35M in just over a year. HeyGen makes visual storytelling accessible by allowing users to create, localize, and personalize studio-quality videos without the need for cameras, actors, or crew.
Decagon Raises $35M Series A: The gen AI enterprise customer support company emerged from stealth today and announced its $5M Seed (a16z) and $30M Series A (Accel) funding rounds.
OpenAI acquires Rockset and Multi: OpenAI’s first acquisition, Rockset, is a leading real-time analytics database company and will be used to power OpenAI's retrieval infrastructure. Multi is a video-focused remote collaboration platform. that could potentially lead to new remote collaboration features in OpenAI's products.
Ilya Sutskever, OpenAI’s former chief scientist, is starting a new AI company: Safe Superintelligence Inc.’s (SSI) primary focus is to develop a safe and powerful AI system, without the distractions of commercial pressures or product cycles. This comes in the wake of Sutskever's departure from OpenAI, which followed a period of internal conflict, including his involvement in the attempt to oust CEO Sam Altman in 2023.
Google’s new startup program focuses on bringing AI to public infrastructure
Tech 💻
Google DeepMind Unveils a New AI model V2A that can Generate Soundtrack and Dialogues for Videos (by Watching Them!): V2A can create synchronized audio, soundtracks or sound effects for AI-generated videos using just the video pixels and a text prompt for additional guidance.
Runway Introduces Gen-3 Alpha: The latest advancement in AI-driven video generation and a significant leap from its predecessor, Gen-2. It can handle both images and videos simultaneously, resulting in improved realism and fidelity. Runway has partnered with leading entertainment and media organizations to create custom versions of Gen-3.
NVIDIA releases Nemotron-4 340B for Synthetic Data Generation: Synthetic data can be used in conjunction with real-world data to enhance AI model training, especially in scenarios where real data is limited or privacy concerns restrict data usage.
Not all ‘open source’ AI models are actually open: here’s a ranking: Many of the LLMs that power chatbots claim to be open, but restrict access to code and training data.
LangChain’s Harrison Chase on Building the Orchestration Layer for AI Agents
Regulation and Policy ⚖️
Apple Won’t Roll Out AI Tech In EU Market Over Regulatory Concerns: They plan to withhold Apple Intelligence from EU this year.
Major Record Labels Sue AI Music Generators Suno and Udio: The record labels (Universal, Sony, Warner) argue that these AI platforms "saturate the market with machine-generated content that will directly compete with, cheapen, and ultimately drown out the genuine sound recordings". They seek damages of up to $150K per infringed work.
Instagram Photos Are Being Labeled ‘Made With AI’ When They’re Not