GSV's AI News & Updates (07/17/24)

Andrej Karpathy's Eureka Labs, LAUSD Ed, Mary Meeker on AI and Higher Ed, Instructure's AI Releases, US DoE AI Guide, OpenAI "Strawberry" Model

Claire Zau

Jul 17, 2024

OpenAI is building a new AI model under code name ‘Strawberry’: This project is said to be focused on enabling AI to plan ahead and perform more complex problem-solving tasks. The project previously went by the name of Q* (pronounced “Q star”), demos of which showed earlier this year that it could answer “tricky science and math questions”.
OpenAI and Arianna Huffington are working together on an ‘AI health coach’: The AI-powered health coach aims to provide personalized wellness recommendations and assist in chronic disease prevention. It will utilize biometric data and medical information to offer tailored guidance on various health domains, emphasizing behavior change through small habit adjustments.
OpenAI Develops System to Track Progress Toward Human-Level AI: They believe their technology is approaching the second level of five on the path to AGI.
Amazon’s AI shopping assistant rolls out to all users in the US: Rufus is designed to help customers save time and make more informed purchase decisions by answering questions on a variety of shopping needs and products.
AI’s $600B Question: The AI bubble is reaching a tipping point. Navigating what comes next will be essential.
- The AI summer: Maybe we have to go through the slow, boring hunt for product-market fit after all.

Education and the Future of Work 📚

Andrej Karpathy, former head of AI at Tesla and researcher at OpenAI, is launching Eureka Labs, an “AI native” education platform: Karpathy's vision for Eureka Labs is to create a symbiosis between human teachers and AI, potentially running an entire curriculum of courses on a common platform. This approach aims to address the scarcity of expert tutors while leveraging the recent progress in generative AI to provide personalized, on-demand learning experiences. Their first product will be an AI course, LLM101n.
Turmoil Surrounds LA’s New AI Student Chatbot as Tech Firm Furloughs Staff Just 3 Months After Launch: Ed was touted as a powerful, easy-to-use online tool for students and parents to supplement classroom instruction, find assistance with kids’ academic struggles and help families navigate attendance, grades, transportation and other key issues, all in 100 languages and on their mobile phones.
- An Education Chatbot Company Collapsed. Where Did the Student Data Go?
- Supt. Carvalho moves ahead with troubled AI effort despite collapse of tech contractor
Mary Meeker’s AI and Higher Education Report: Mary Meeker has written her first report in over four years, focused on the relationship between AI and U.S. higher education. Report here.
New US Department of Education AI Guide Launched for Edtech Developers: New guidance recommends that AI not be allowed to make decisions unchecked by educators, and that developers design AI tools based on evidence-based practices. Guide here.
Instructure Announces Release of AI Features: The new AI product announcements included features developed directly within Instructure products, a number of well-tested partner integrations, including Khan Academy's Khanmigo Teacher Tools and updates that enable partners to develop context-aware AI experiences within Canvas. The company also announced the launch of a standalone product, Instructure Intelligent Insights.
Imagine Learning Introduces AI Grading Assistant for Writing in Imagine EdgeEX: The AI Grading Assistant analyzes student writing submissions and offers suggestions on areas such as clarity of thought, alignment to learning objectives, and overall writing quality.
New Guidance from TeachAI and CSTA Emphasizes Computer Science Education More Important than Ever in an Age of AI: 52% of teachers believe the benefits of AI in CS education outweigh the risks. 80% of teachers agree that core CS concepts should be updated to better support learning. 88% of teachers need additional resources and professional learning to teach with and about AI.
AI Work Assistants Need a Lot of Handholding: Getting full value out of AI workplace assistants is turning out to require a heavy lift from enterprises. ‘It has been more work than anticipated,’ says one CIO.
At Target, Store Workers Become A.I. Conduits: The retailer is rolling out a chatbot to help workers answer questions from shoppers — and workers.
What teachers call AI cheating, leaders in the workforce might call progress: “A core tension has emerged: many teachers want to keep AI out of our classrooms, but also know that future workplaces may demand AI literacy.
What we call cheating, businesses could see as efficiency and progress.”
70,000 students are already using AI textbooks: Pearson has integrated AI-powered features into 50 science textbooks, now used by 70,000 students across over 1,000 institutions, offering interactive chatbots and personalized practice questions to enhance learning experiences.
Homeschoolers Embrace AI, Even As Many Educators Keep It at Arms’ Length

Startups and Tools 🛠️

Hebbia raises $130M Series B led by a16z: Hebbia utilizes generative AI to search large documents and respond to complex questions. Hebbia's customers include leading asset managers, law firms, banks, and Fortune 100 companies. The funding round values Hebbia at approximately $700M.
Exa raises $22M Seed and Series A to build the Search Engine for AI: The funding round was led by Lightspeed Venture Partners, with participation from NVentures (NVIDIA's venture capital arm) and YC. Exa's CEO envisions a future where AI will search the web more frequently than humans, necessitating a search engine specifically designed for AI needs.
Medal raises $13M as it builds out a contextual AI assistant for desktop: The company also unveiled Highlight. The app captures the content on your screen and lets you ask questions to an LLM based on that context.
Sierra AI: Allows businesses to deploy AI agents that are available 24/7, empathetic, and aligned with the company's brand tone and voice.
Investors Pour $27.1 Billion Into A.I. Start-Ups, Defying a Downturn
Goldman Sachs: Gen AI: too much spend, too little benefit?: Tech giants and beyond are set to spend over $1tn on AI capex in coming years, with so far little to show for it. So, will this large spend ever pay off?
a16Z Is Building a Stash of More Than 20,000 GPUs to Win AI Deals: The initiative is aptly named Oxygen, because these chips are that integral to AI companies. The chips are almost impossible to secure for small startups too, because Big Tech companies hoover up all the supply.

Tech 💻

Microsoft CTO Kevin Scott thinks LLM “scaling laws” will hold despite criticism: Microsoft CTO Kevin Scott reaffirmed his belief in the continued effectiveness of scaling laws for improving LLMs, aligning with OpenAI's 2020 research suggesting that increasing model size, training data, and computational power can significantly enhance AI capabilities without major algorithmic breakthroughs. This optimism contrasts with a prevailing narrative in the AI community that suggests progress in models similar to GPT-4 may be slowing down.
Meta to reportedly launch largest Llama 3 model on July 23: The new model, boasting 405 billion parameters, will be multimodal and capable of understanding and generating both images and text.
Meta shows off ‘3D Gen’ AI tool that creates textured models faster than ever
Anthropic Pushes for Third-Party AI Model Evaluations: This initiative seeks to enhance the AI safety ecosystem by funding the creation of high-quality evaluations from third-party organizations.
Patronus AI Launches Lynx: State-of-the-Art Open Source Hallucination Detection Model: Lynx outperforms existing solutions like GPT-4, Claude-3-Sonnet, and other LLM-as-judge models in detecting hallucinations. Lynx and HaluBench are now publicly available on Hugging Face for researchers and developers to use and contribute to.
Extrinsic Hallucinations in LLMs

Regulation and Policy ⚖️

Apple, Nvidia, Anthropic Used Thousands of Swiped YouTube Videos to Train AI: An investigation by Proof News revealed that these companies utilized subtitles from 173,536 YouTube videos, sourced from over 48,000 channels, despite YouTube's policies prohibiting unauthorized data extraction. Content creators and platform owners are now grappling with the implications of their work being used without permission or compensation for AI training purposes.
Microsoft and Apple ditch OpenAI board seats amid regulatory scrutiny: This move comes as a response to increasing antitrust concerns surrounding the influence of major tech companies in the artificial intelligence sector.