GSV's AI News & Updates (06/09/24)
GPT-4 Interpretability, Google AI Overview Flaws, xAI raises $6B, Doowii raises $4.1M, Alibaba Invests $27M in AI EdTech, NotebookLM, EDSAFE AI
General 🚀
Extracting Concepts from GPT-4: Following Anthropic’s breakthrough last week, OpenAI has also developed new scalable methods to decompose GPT-4's internal representations into 16 million interpretable patterns or "features." Both are promising steps towards increasing the interpretability and trustworthiness of LLMs.
Google Rolls Back AI Search Feature After Flubs and Flaws: In May 2024, Google unveiled AI Overviews. The feature quickly ran into issues, providing flawed information like suggesting glue as an ingredient for pizza recipes and advising people to consume rocks for nutrients.
Google AI Overviews visibility drops, only showing for 15% of queries: (formerly known as Search Generative Experience when it was an opt-in experiment in Google Labs) at one time appeared on 84% of queries.
Elon Musk’s xAI raises $6B from Valor, a16z and Sequoia: xAI raised $6B in funding, valuing the company at $24B post-money. They launched their flagship product, the AI chatbot Grok, in November 2023 as a rival to OpenAI's ChatGPT. Musk aims for it to be "truth-seeking" and open source.
The CEO of Zoom wants AI clones in meetings: Eric Yuan envisions a future where AI-powered "digital twins" can attend meetings on your behalf, allowing you to spend more time on other activities.
Media Companies Are Making a Huge Mistake With AI: News organizations rushing to absolve AI companies of theft are acting against their own interests.
How Apple Fell Behind in the AI Arms Race
What to expect from Apple’s AI-powered iOS 18 at WWDC 2024: Siri’s AI-enabled future, Apple Photos’ new look, Safari with more AI, iMessage gets an AI makeover, and more.
The AI Revolution Is Already Losing Steam: While big tech is snapping up AI chips, the excitement might be premature as AI improvements slow, costs soar, and practical uses remain limited.
The New ChatGPT Offers a Lesson in A.I. Hype: OpenAI released GPT-4o, its latest chatbot technology, in a partly finished state. It has much to prove.
Education and the Future of Work 📚
Doowii secures $4.1M to advance AI-driven data analytics for educators: Doowii is a pioneering EdTech company focused on enhancing educator access to advanced data analytics and serving as a modern data interoperability layer for various EdTech data sources. By providing powerful tools for data analysis and visualization, Doowii empowers educators to make informed decisions and boost student performance.
Alibaba invests US$27 million in education start-up to expand its generative AI portfolio amid price war in China: The investment aims to support the development and promotion of Jingzhunxue's AI-powered education device, Bong, which offers AI-generated video lessons taught by digital teachers. Bong is powered by Alibaba's Qwen model and Jingzhunxue's self-developed LLM.
Imagine Learning Acquires CueThink: CueThink is an AI-powered problem-solving and collaboration platform designed to enhance critical thinking skills in K-12 education.
Google’s NotebookLM goes global with Slides support and better ways to fact-check: New features include global availability, support for Google Slides and websites as sources, inline citations for fact-checking, multimodal capabilities to understand images/charts, and a "Notebook Guide" feature to generate overviews.
50+ Education and Tech Companies Make Commitment to Responsible AI Development: The EDSAFE AI Alliance recently launched its EDSAFE AI Industry Council, comprising over 50 pioneering education and technology companies. The council aims to ensure artificial intelligence (AI), generative AI, and related technologies within their organizations and across the education sector are safe, accountable, fair, and effective (SAFE).
Say Hello to My New AI Marketer: How Gen AI-Based Software Is Advancing Marketing and Sales
Over half (51%) of young people aged 14–22 have tried generative AI, but only 4% use it daily, and 8% haven't heard of it.
Black and Latinx youth use generative AI more frequently than White youth and for a wider range of activities.
Young people are excited about AI's potential to boost efficiency and creativity but worry about job loss, intellectual property theft, personal data misuse, and misinformation.
Can I Use AI to Grade My Students’ Papers?: NYT Ethicist columnist on artificial intelligence platforms, and whether it’s hypocritical for teachers to use these tools while forbidding students from doing the same.
AI essay grading is already as ‘good as an overburdened’ teacher, but researchers say it needs more work: Early research by Tamara Tate and her team at UC Irvine shows that ChatGPT can grade essays nearly as accurately as a busy teacher, suggesting it could help alleviate this burden. While ChatGPT’s accuracy is promising, its use should be limited to low-stakes grading until it improves.
Latest AI Announcements Mean Another Big Adjustment for Educators
PowerSchool to be Acquired by Bain Capital in $5.6B Transaction
Reid Hoffman and Aria Finger interview Pat Yongpradit on AI literacy and computer science education (live podcast from ASU+GSV!)
Startups and Tools 🛠️
Doing Stuff with AI: Ethan Mollick provides an updated guide on using advanced AI systems like Claude 3 Opus, Gemini 1.5, and GPT-4 for practical applications. He advocates to "always invite AI to the table" for tasks like getting advice, summarizing meetings, generating ideas, writing, producing reports, and discussing strategy.
French AI startup H raises $220M seed round: to develop advanced "frontier action models" or "agentic AI". These models aim to boost worker productivity by automating complex tasks through multi-step reasoning, planning, and collaboration capabilities that surpass current AI assistants. Four of the five co-founders all previously worked for DeepMind.
Pika raises $80M Series B: Pika's tool allows users to generate short videos from text prompts, images, or existing video clips.
Coda introduces Coda Brain - your team’s favorite know-it-all: Coda is launching a turnkey AI platform that helps teams spend less time hunting for information—and more time putting data to work.
Perplexity Pages: a simple way to turn your research into visually appealing articles.
Sirion: enterprise AI for contracts.
Telly: AI video stories.
Tech 💻
AI in software engineering at Google: Progress and the path ahead: Google is integrating its latest Gemini series foundation models into developer tools to enhance productivity, with AI projected to bridge gaps in mental health support, fault detection, code optimization, team interaction, and skill learning for developers by 2030.
FineWeb: decanting the web for the finest text data at scale: A massive dataset of 15 trillion tokens derived from CommonCrawl web snapshots, designed for pretraining large language models (LLMs).
Regulation and Policy ⚖️
OpenAI Insiders Warn of a ‘Reckless’ Race for Dominance: A group of current and former employees is calling for sweeping changes to the artificial intelligence industry, including greater transparency and protections for whistle-blowers.
OpenAI Shares Insights on How It's Securing AI Research Infrastructure: OpenAI has implemented a comprehensive security architecture leveraging Azure identity services, Kubernetes isolation, strict access controls, encrypted data storage, secure CI/CD pipelines, defense-in-depth for model protection, and rigorous testing to safeguard its cutting-edge AI research infrastructure against threats.
Microsoft is reworking Recall after researchers point out its security problems: Recall was initially designed to provide a searchable timeline of a user's activities on their PC, like a personal time machine. Cybersecurity researchers say the system is simple to abuse—and one ethical hacker has already built a tool to show how easy it really is.
Indian Voters Are Being Bombarded With Millions of Deepfakes. Political Candidates Approve: India’s elections are a glimpse of the AI-driven future of democracy. Politicians are using audio and video deepfakes of themselves to reach voters—who may have no idea they’ve been talking to a clone.
Other
Vinod Khosla, Marc Andreessen And The Billionaire Battle For AI's Future
Eric Schmidt Is Secretly Testing AI Military Drones: The secretive venture, previously known as White Stork and now reportedly called Project Eagle, aims to create drones that leverage AI technology to accurately identify and strike targets on the battlefield.
Resources
The GSV BIG 10: weekly coverage corner for the top 10 stories, insights, and major plays in learning and skilling.