GSV's AI News & Updates (10/01/24)
OpenAI Leadership Exits, Advanced Voice Mode, Meta Orion, Cerebras IPO, Duocon, Workera AI Agent, Udemy AI Training, Newsom Vetoes SB1047, Sovereign AI, OpenAI Financials
General 🚀
OpenAI CTO Mira Murati is leaving: Murati announced her departure after six and a half years, stating she needs time for personal explorations. Her exit follows a series of leadership changes at the company - Murati is one of more than 20 senior execs, researchers and co-founders who have quit this year.
OpenAI to remove non-profit control and give Sam Altman equity: This move could raise concerns about AI safety governance but aligns OpenAI's structure with competitors like Anthropic and xAI, while also giving CEO Sam Altman equity for the first time, potentially valuing the company at $150B.
OpenAI is reaching $300M in monthly revenue in August and projecting $3.7B for 2024, with plans to hit $11.6 billion next year.
Despite significant revenue, OpenAI anticipates a loss of approximately $5B this year due to high operational costs, including expenses related to computing power.
It has 350M monthly active users (up from 100M in March).
Approximately 10M users pay a $20 monthly fee, with plans to increase the price up to $44 over the next five years.
OpenAI rolls out Advanced Voice Mode: This update includes five new voices, improved accent understanding, and enhanced conversation flow. This feature is currently being rolled out to Plus and Team subscribers, with a daily usage limit of 30-120 minutes (depending on the day and user feedback).
Meta offers a glimpse through its supposed iPhone killer: Orion: Meta's Orion smart glasses prototype features a lightweight design with advanced technology, including microprojectors and eye-tracking capabilities, allowing users to seamlessly overlay digital content onto their physical surroundings. Meta's introduction of Orion, aiming for a consumer version by 2027, puts pressure on competitors like Apple's Vision Pro headset and marks a bold step toward replacing smartphones with wearable tech.
EA Shows Off AI System That Lets You Dream Up an Entire Videogame From Scratch: "Imagination to Creation" allows users to design entire video games from scratch using simple prompts, sparking both excitement for creative possibilities and concerns about job security in the gaming industry.
AI chipmaker Cerebras files for IPO to take on Nvidia: Founded in 2016 and based in Sunnyvale, CA, Cerebras specializes in designing AI chips and offering cloud-based services. Despite reporting a net loss of $66.6M in the first half of 2024, Cerebras secured significant sales from AI firm G42. It faces competition from tech giants like Microsoft, Google, and AMD.
Education and the Future of Work 📚
Duolingo Introduces AI-Powered Innovations at Duocon 2024:
AI-Powered Conversations: Video Call with Lily – Duolingo's new Video Call feature allows Duolingo Max subscribers to practice real-time, AI-powered conversations with the character Lily, offering personalized and adaptable language practice.
Adventures: Explore and learn in the Duolingo World – Duolingo's Adventures feature immerses learners in a game-like exploration experience, enabling them to interact with characters and practice language skills in realistic, dynamic scenarios.
Expanding Music: Popular songs, compact piano partnership, and an A-list interview – Duolingo has partnered with Loog to create a compact digital piano for music lessons, and with Sony Music to offer learners the chance to play along to popular songs.
Workera’s CEO was mentored by Andrew Ng. Now he wants an AI agent to mentor you: Sage provides personalized responses for individuals and organizations, adapting to their specific needs, goals, and skill levels. Learners can receive guidance and motivation from the AI mentor, while managers can leverage Sage to assess talent and receive recommendations for enhancing workforce development.
AI reading coach startup Ello now lets kids create their own stories: The AI helps with reading skills by listening to kids read aloud, correcting mispronunciations, and adapting stories to their reading level, while also incorporating phonics-based strategies.
How agentic AI platforms will redefine enterprise applications: These systems will allow armies of agents to work together, bridging legacy application silos and boosting productivity, potentially reducing the need for human intervention in common business tasks by up to 90%, while creating a new intelligent application stack over the next decade.
I Quit Teaching Because of ChatGPT: A longtime university instructor, the author quit teaching because of students' growing reliance on AI tools like ChatGPT, which prevented them from fully engaging in the cognitive process of writing and critical thinking.
AI is an existential threat to colleges. Can they adapt?: Colleges face a dilemma: embrace AI and adapt coursework, or return to traditional methods like handwritten exams, both of which present challenges in maintaining educational quality and relevance in the AI era.
AI Can (Mostly) Outperform Human CEOs: In an automotive industry simulation, AI models outperformed humans in market share and profitability but struggled with unpredictable disruptions, leading to quicker dismissals by virtual boards. While AI excels at data analysis and rapid iteration, it lacks the intuition to handle black swan events.
E-Learning Platform Udemy Gave Teachers a Gen AI 'Opt-Out Window'. It's Already Over: Teachers are surprised they have been opted into having their classes scraped for AI training. Instructors have voiced frustration over the lack of transparency and difficulty in opting out, citing concerns about intellectual property and course visibility. Despite Udemy's promises of revenue sharing and no penalties for opting out, many feel misled and some have left the platform in protest.
Roombr Aims to Reach 1 Million Classrooms by 2026 with AI-Powered Digital Classroom: In collaboration with Intel and Texas Instruments, the system integrates interactive screens and personalized learning tools, and has already been adopted in over 3,000 classrooms.
Startups and Tools 🛠️
11x.ai, a developer of AI sales reps, has raised $50M Series B led by A16Z: The company, which previously raised $24M in Series A, offers AI sales development bots named Alice (for lead generation and outreach) and Jordan (a multilingual AI phone sales rep), and is nearing $10M in ARR, positioning itself among competitors in the AI SDR space.
Arcade, the world's first AI product creation platform, raises $17M: The platform is currently in beta and allows users to generate physical products simply by typing a few words or uploading images, making the process of product creation accessible and engaging.
Tech 💻
Google rolls out two new Gemini models: Gemini-1.5-Pro-002 and Gemini-1.5-Flash-002 offer developers a 50% price drop, two times faster output, and three times lower latency. These models, designed for tasks like code generation, text processing, and multimodal tasks, feature improved performance, streamlined access, and updated privacy filters.
Google unveils AlphaChip AI-assisted chip design technology: AlphaChip uses reinforcement learning to create superhuman chip layouts in a matter of hours, dramatically accelerating a process that traditionally took weeks or months of human effort. AlphaChip approaches chip floorplanning as a game-like task.
Safety and Regulation ⚖️
California’s governor vetoes major AI Safety bill ’SB1047’: Citing concerns that the bill's broad scope could hinder innovation and give a false sense of security. Despite acknowledging the need for AI safety protocols, Newsom argued that the legislation imposed overly stringent standards on AI companies without considering the specific risks posed by different systems.
SB 942 requires generative AI providers with over 1M monthly users to offer free AI detection tools to users and include "latent disclosures" in AI-generated content's metadata. The law takes effect on January 1, 2026, and violations can result in civil penalties of $5,000 per day.
SB 926 and SB 981 targets the misuse of AI for creating and distributing non-consensual sexually explicit content.
Why countries are in a race to build AI factories in the name of sovereign AI: Nvidia is a major player in supporting sovereign AI, investing $110 million in AI startups and collaborating with countries to build "AI factories." The push for sovereign AI is seen as a way for nations to control their AI resources, much like the industrial revolution, while addressing issues like energy consumption, security, and privacy.
OpenAI Training Data to Be Inspected in Authors’ Copyright Cases: For the first time, OpenAI will provide access to its training data for review of whether copyrighted works were used to power its technology.
Other
James Cameron is joining Stability AI’s board of directors: Cameron’s move comes as studios increasingly collaborate with AI firms, though these partnerships have sparked controversy, contributing to Hollywood's recent strikes and raising concerns over the use of AI in content creation.
Spotify’s AI playlist builder is now available in the US: AI playlists can now be created on the Spotify mobile app by subscribers in the US, Canada, Ireland, and New Zealand.
‘Three New York Cities’ Worth of Power: AI Is Stressing the Grid: Across the nation, utilities are worried about expanding the overburdened power grid, citing high costs and concerns about commitment from data center projects.
