AI Weekly Update Lecture Notes

Introduction

• Busy week in AI
• Numerous topics to cover
• Video might be longer than usual

OpenAI's Advanced Voice Feature

• Rolling out new advanced voice feature (e.g., sounds like Scarlett Johansson)
• Limited access, not widely available yet
• Features demonstrated:
  • Voice acting as a frog, airline pilot, counting fast, etc.
  • Users can interrupt while it speaks

GPT-4.0 Long Output

• Experimental version with 64,000 output tokens per request
• Available in API to Alpha participants

Microsoft Labels OpenAI as Competitor

• Despite owning 49% of OpenAI, Microsoft lists it as a competitor in AI and search
• Financial reports highlight competition with Google, Meta, and open-source offerings

OpenAI Endorses Senate Bills

• Future of AI Innovation Act: Authorizes U.S. AI Safety Institute
• NSF AI Education Act and Create AI Act: Scholarships and educational resources
• OpenAI aiming for regulatory goodwill and influence

U.S. Government and AI Regulation

• White House: No need to restrict open-source AI currently
• Concerned about slowing progress and competition with other countries

News from Anthropic

• Claude AI available in Brazil

Qualcomm Innovations

• AI advancements in AR glasses, facial expression input, deep fake detection, biometric palm scanner, AI-powered DJ software, and more

Google Updates

• Released Gemini 1.5 Pro (version 0801)
• Topped LM leaderboard
• Released a smaller 2B parameter model outperforming larger models
• New Chrome AI features: Google Lens, compare products, search history with natural language

Meta Updates

• Discontinued AI chatbots featuring famous faces
• Launched AI Studio for custom AI characters
• Released SAM 2 (Segment Anything Model) with improved segmentation capabilities

Perplexity AI Publishers Program

• Revenue sharing with specific partner publishers
• Initial partners: Time, Der Spiegel, Fortune, Entrepreneur, Texas Tribune, WordPress.com

Leonardo AI Acquired by Canva

• Leonardo AI will still operate independently
• Integration will enhance Canva's AI capabilities

MidJourney Update

• Version 6.1 improves image quality, text coherence, new upscaling, and personalization model

3D Model Generators

• Nvidia and Shutterstock's edify 3D
• Stability AI's stable Fast 3D
• Black Forest Labs launched Flux model

Runway ML Updates

• Released image-to-video model
• Announced Gen 3 Alpha turbo for faster, cheaper video generation

AI Character Creator Apps

• Render.net's Narrator: Create AI-generated lip-synced videos
• Captions' AI Avatar: Digital twin creation

Vimeo AI Translation

• Translates videos using your own voice into different languages

Suno's Legal Issues

• Responded to lawsuits regarding training on copyrighted information

Friend AI Device Drama

• New “Friend” device listens all day, sends text messages
• Controversy over similar product and domain purchase

Video Game Performers Strike

• Concerns over AI replicating voices and likenesses without consent

AI in Fast Food

• Taco Bell to implement AI in drive-throughs by end of 2024

Miscellaneous AI Innovations

• AI toothbrush by ARS Technica
• AI technologies in the Olympics for better judging and analysis

Conclusion

• Anticipation of more AI advancements and updates
• Personal reflection on content creation and focus on exciting AI topics