• Future Perfect
  • Posts
  • #116. 🇪🇺 EU AI Act deep dive 💰 Former OpenAI CTO raising $100M for new startup 🩻 UCLA's AI medical imagery breakthrough

#116. 🇪🇺 EU AI Act deep dive 💰 Former OpenAI CTO raising $100M for new startup 🩻 UCLA's AI medical imagery breakthrough

Plus, AI tools: Granola: AI note taking; Graphy: AI-powered charts; Descrybe: free AI legal research

🎙️ Check out my new Marc Hoag AI Law Pod, a podcast “hosted” by Google's NotebookLM. Now that NotebookLM allows customization, the AI “hosts” actually interact with me, directly reference the newsletter itself, and deep dive quite a bit beyond the bounds of each issue’s content. If nothing else, it’s a very cool alternative way to enjoy Future Perfect while driving.

  • The EU AI Act introduces a tiered risk-based approach to AI regulation, categorizing systems from prohibited to minimal-risk.

  • High-risk AI systems face strict requirements, including compliance checks, risk management, and thorough documentation.

  • The Act specifically addresses transparency in AI, covering areas like deep fakes and general-purpose AI.

  • Non-compliance can lead to severe penalties, with fines reaching €35 million or 7% of global revenue.

  • The Act aims to ensure AI innovation while safeguarding public safety and fundamental rights.

  • Mira Murati, the former CTO of OpenAI, has started fundraising for her own AI startup after departing the company in late September 2024.

  • The new venture is expected to focus on developing proprietary AI models, with Murati reportedly recruiting talent from OpenAI for the project.

  • Murati's departure is one of several high-profile exits from OpenAI, which is undergoing governance changes, including the shift from a non-profit to a for-profit structure.

  • Barret Zoph, another senior researcher from OpenAI, is rumored to be involved with Murati's new venture, though details remain speculative.

  • This trend reflects a broader movement, with several former OpenAI employees launching their own AI startups, including notable figures like Ilya Sutskever and Andrej Karpathy.

  • Researchers at UCLA developed SLIViT, an AI model that analyzes 3D medical images rapidly and cost-effectively.

  • It uses deep learning across multiple imaging types, identifying disease biomarkers, often outperforming existing models.

  • The model’s 2D pre-training approach allows it to excel at 3D analysis, reducing costs and resource demands.

  • NVIDIA GPUs like the T4 and V100 were used to power the model’s development.

  • SLIViT’s scalable and adaptable nature could improve patient outcomes, particularly where medical expertise is limited.

✨ PowerPrompt™

See attached PDFs for how [xyz] is supposed to work. I want to make an app for this. I'll be using Cursor AI or Replit, so generate a prompt that I can use that will enable those tools to build an MVP for me.

If ever there was an example of “YMMV” — your mileage may vary — this is certainly it. But the idea is that you can use AI — in this case either ChatGPT or Claude — to imagine building a tool based on a set of PDFs you give it, and to then generate a prompt to help you get another AI development tool to build it for you. When it works — which is usually, more or less anyway — it’s amazing.

🧰 AI Tools & Resources

  • Granola AI: Fleshes out your fragmented meeting notes.

  • Graphy: Generate beautiful AI-powered charts.

  • Descrybe: Free AI-powered legal research

🎉 THAT’S ALL FOR TODAY!

See you next time! 👋

What'd you think of today's email?

This helps me improve it for you!

Login or Subscribe to participate in polls.

Looking for past newsletters? You can find them all here.

Reply

or to participate.