Upcoming Events
SF & North Bay
Thu, July 11th: 🧠 GenAI Collective 🧠 The Diversity AI Forum
Thu, July 18th: 🧠 GenAI Collective 🧠 AI Video Quest 📽️
Thu, July 18th: 🧠 The GenAI Collective 🧠 Office Hours 👨🏫 (Virtual)
Tue, July 23rd: 🧠 GenAI Collective 🧠 Garden Gathering 🌺
Tue, Aug 13th: 🧠 GenAI Collective 🧠 Research Roundtable 🧑🔬
🗓️ Hungry for even more AI events? Check out SF IRL, MLOps SF, or Cerebral Valley’s spreadsheet!
NYC
Tue, July 9th: GenAI Collective NYC Presents: SWE-Agent & the Future of Software Engineering
Wed, July 17th: GenAI Collective Panel Discussion @ The Inside Summit
The GenAI Collective is thrilled to have the NYC community join us for an exclusive panel discussion at the Inside Summit on July 17th, in partnership with The Room Podcast. Join us for an insightful conversation featuring distinguished guests: Dynamo AI CEO Vaikkunth Mugunthan, Tavus CEO Hassaan Raza, and Multimodal CEO Ankur Patel. Together, we’ll explore the current state of Generative AI and delve into what it takes to build a sustainable and differentiated business in this rapidly evolving field. Don’t miss this opportunity to gain valuable insights from industry leaders. Learn more here.
CA SB-1047: The Scoop on California’s New AI Compute Law
On May 21st, 2024, the California Senate passed Bill 1047, also known as the "Safe and Secure Innovation for Frontier Artificial Intelligence Systems Act,” via an overwhelming bipartisan vote of 32 to 1. This bill is set for an assembly vote in August that would enact it into law. Given its implications on the future of responsible AI development, we encourage everyone to read the bill and discuss the near and long-term impact on one of the most transformation technologies of our lifetime if enacted.
The Gist of SB-1047
SB-1047 is set to create a new standard for AI development, and is one of the first legislative initiatives to target the AI model layer–most of the initial AI legislation over the last 12 months was focused on more specific applications (deepfakes, AI systems in employment, data privacy). Not all AI models are scrutinized equally under the Act. The bill focuses specifically on models that either use more than 1026 floating-point operations (FLOPs) during training or “has similar performance to that of state-of-the-art foundation models,” aiming to prevent the most powerful systems from posing existential threats. The bill also establishes the Frontier Model Division within the Department of Technology for oversight of these AI models, introducing civil penalties for violations of the Act.
Key Requirements:
Positive Safety Determination: Before starting training, developers must certify that their AI models do not have hazardous capabilities, such as creating weapons of mass destruction or causing significant cyber damage. This certification must be submitted to the newly created Frontier Model Division under penalty of perjury (Center for Data Innovation) (Benesch & FC&A).
Full Shutdown Capability: AI models must include mechanisms for a complete shutdown, ensuring that all copies and derivatives can be promptly deactivated if needed (LegiScan) (Benesch & FC&A).
Compliance with Guidance: Developers must adhere to guidelines from the National Institute of Standards and Technology (NIST) and follow industry best practices to ensure the safety and security of their AI models (Center for Data Innovation) (Benesch & FC&A).
Third-Party Model Testing: Developers of these AI models are required to implement a written safety and security protocol which provides enough detail for third parties to test the safety of the models they bring to market (DLA Piper).
Annual Reviews: Continuous evaluation and updates of safety protocols are mandatory to keep up with the evolving landscape of AI technology (Benesch & FC&A).
Whistleblower Protections: The bill incorporates provisions to protect and encourage whistleblowing within AI companies to ensure the most equipped employees can report non-compliance without fear of retaliation (DLA Piper).
Contextualizing the Meaning of “1026 Flops”
GPT-3: Uses about 1023 FLOPs. Known for generating human-like text and performing various tasks.
AlphaGo Zero: Utilizes around 1026 FLOPs, excelling in the game of Go by defeating human champions.
Future AI with 1026: These models could be 1000x more powerful than GPT-3, potentially approaching general intelligence and handling a wide array of tasks with unprecedented sophistication.
Key Implications:
Competitive Disadvantage: California-based AI companies might face hurdles that out-of-state or international competitors don't, potentially slowing their progress.
Startup Struggles: Smaller AI firms could find it tough to meet all the new requirements, possibly giving big tech an even bigger edge (Wired) (Forbes).
Potential AI Exodus: Some companies might pack up and leave California to dodge these regulations, potentially causing a "brain drain" in the state's AI sector
Efficiency Push: The FLOP focus might inadvertently spark a race for more efficient, powerful models that fly under the regulatory radar (ArXiv) (IEEE Spectrum).
Safety-First Shift: We could see a pivot towards more thoughtful, safety-oriented AI development practices (Nature) (AI Ethics Journal).
Why It Matters?
SB-1047 isn't just about keeping AI in check in California; it's setting the stage for AI regulation across the country. This bill aims to balance cutting-edge AI development with responsible innovation. While it is creating a relatively high bar for regulation to avoid stifling innovation, it's clear that AI safety is now a top priority for lawmakers. Going forward, both AI Researchers and their employers will be on the hook for the tech they create. New testing standards are bound to pop up in the next year or so, and companies of all sizes will need to get serious about meeting these new safety requirements.
Whether you think this bill will put the brakes on tech progress or is a crucial step towards safer next-gen AI, there's no denying that California is positioning itself as a pioneer in responsible AI development with SB-1047. This could either become the gold standard for balancing AI innovation and safety or push top AI talent and companies to greener pastures. Either way, we're at a turning point in how we govern one of the most transformative technologies of our time!
Some Public Commentary on SB-1047
Acting as a petition that’s picked up traction, “StopSB1047” (funded by a16z), AI developers, researchers and founders Senate Bill 1047 is criticized for potentially stifling AI innovation in California, harming small businesses, and giving other countries, like China, an advantage in AI development. Experts argue that the bill's vague definitions and stringent regulations could create significant economic and security risks, urging lawmakers to reject it.
Jeremy Howard raised concerns about the potential for over-regulation. He believes that while safety and ethical considerations are essential, overly stringent regulations might stifle innovation and hinder the competitive edge of AI companies in California. Howard advocates for a balanced approach that promotes both innovation and safety.
The Gen AI Collective would love to hear what the community thinks about the introduction of this new legislation. If you would like to share your thoughts, hit up Eric on the community Slack, or reach out via email at eric@gaicollective.com for a feature in our next newsletter! For a deeper dive into SB-1047, check out the full details at the Center for Data Innovation, LegiScan, and Benesch Law.
Events Recap
We took it to the NEXT LEVEL with Thursday’s SF Demo Spectacular! Following an exclusive deep tech VC panel led by Lisa Yu to kick us off, we had 50+ early-stage AI companies apply for the chance to ENTER THE ARENA 🏟️, demoing to and receiving feedback from 250+ attendees!!
🧑💻 Featured Companies
AUR+A Networking: a face recognition and AR app for event organizers and attendees (Josephine Declerck)
LearnQuantum: AI-native physics engine (Richa Sharma)
Advex AI : Solve Data. Solve AI. (Pedro Pachuca)
just words: Next-gen AI marketing platform to scale hyper-personalized messaging (Neha Mittal)
mutable.ai: Creates codebase documentation and keeps it up to date using AI (Maxwell Siegelman)
Agentplace / Machinet.net: Agentic websites. A website that is a GPT model (Uladzislau Yanchanka)
SpeedLegal: Review contracts and find issues for non-lawyers (Hans Paul Pizzinini)
Commune AI: Deploy and serve ML models in one line of code (Rohan Sreerama)
Andi: Your AI brain for organizing the web and smashing information overload (Jed W.)
MultiOn: Personal AI agents that revolutionize how we interact with the digital world (Div Garg)
At the end of the night, we held a community vote for the best demos to be immortalized in the Demo Night Hall of Fame! (check out demos.genaicollective.ai) After nearly 500 votes, here are the winners! 🗳️
🏅 Award Winners
🎨 Best Design: MultiOn
🤖 Best Technology: Advex AI
🏆 Best Overall: SpeedLegal
This event would not have been the incredible success it was without the active participation of ~literally everyone~ in attendance who submitted nearly 1,000 pieces of invaluable feedback to the demoists! 🤩
Stay tuned for another exciting announcement of our next demo night in mid-August!! 👀
Value Board
🌎 GenAI Collective Coming to You!
Skip the trip to SF because the GenAI Collective is coming to you! We’re looking for the best and the brightest to jumpstart new chapters around the globe. This is your chance to dive deep into the AI community, enrich your network, and showcase your skills. If you’re ambitious, organized, and ready to make a splash in the AI scene, we want to hear from you. Let’s create unforgettable events together!
Interested in starting a new chapter? Check out our Chapter Expansion Starter Pack for more details.
Reach out to stephen@genaicollective.ai or check out some of the other open roles on the team! 😄
About Eric Fett
Eric leads the development of the newsletter and online presence. He is currently an investor at NGP Capital where he focuses on Series A/B investments across enterprise AI, cybersecurity, and industrial technology. He’s passionate about working with early-stage visionaries on their quest to create a better future. When not working, you can find him on a soccer field or at a sushi bar! 🍣