🎩 OpenAI Teases Magical Reveal

PLUS: AI Leaderboard Insights

Welcome back!

Just before Google's big annual I/O event, OpenAI is planning to grab attention by revealing a new "magical" AI-powered product next Monday, right before Google’s showcase. Let’s watch what happens…

Today’s Summary:

  • OpenAI hints at upcoming features

  • LMSys releases new model rankings

  • AI-driven film festival

  • ElevenLabs introduces its music AI

  • Apple deploys M2 Ultra chips in servers

  • 3 new tools


OpenAI Preps New Features Before Google I/O

The Summary: OpenAI is set to announce new AI features on Monday, a day before Google's annual I/O event.

While OpenAI denies launching a search engine or GPT-5, the company hints at unveiling something new and "magical". The announcement timing suggests a strategic move to steal the spotlight from Google's AI showcase.

Key details:

  • OpenAI will announce its AI search product on Monday, May 13th, 10AM PT

  • OpenAI CEO Sam Altman tweeted about unveiling "new stuff" on Monday, denying it's either GPT-5 or a search engine

  • Google's stock dipped on the rumors but recovered after Altman's clarification

  • Perplexity AI, a $1 billion AI startup, already has 10 million users for its AI search offering

Why it matters: This move signals OpenAI's growing ambition to challenge Google's search dominance, leveraging the popularity of ChatGPT. The timing also suggests a strategic attempt to overshadow Google's AI announcements at I/O event.


AI Model Leaderboard Reveals New Category Performance Data

Image: DALL·E

The Summary: The LMSys AI leaderboard recently launched a filter that lets users see model performance by category, revealing new insights on model capabilities across domains. The categories span coding tasks to multilingual queries. LMSys also announced that more fine-grained categories will be coming soon.

New data on Llama3 70B model shows mixed performance, it performs well in English writing but has difficulty with complex prompts and tasks in other languages. This detailed breakdown aids in choosing the right model for specific tasks.

Source: LMSys

Key details:

  • Claude 3 Opus matches ChatGPT 4 when not counting refused answers

  • GPT-4 leads in coding prompts

  • Claude 3 Opus leads in Chinese language

  • Llama 3 70B shines in English writing but falls short on complex and non-English prompts.

  • Gemini 1.5 Pro API Preview scores well across the board

  • Detailed report here

Why it matters: The new category filter on the LMSys leaderboard gives users insights into model performance across specific tasks, allowing them to choose models that best meet their particular requirements, thus improving both performance and efficiency.


AI Films Take Center Stage at Runway's LA Festival

Image: Runway

The Summary: The 2nd Annual International AI Film Festival by Runway AI showed how AI is starting to change filmmaking. The festival featured films that mixed AI visuals and effects with solid stories.

The winner, Get Me Out, used emotional live-action scenes and AI-generated settings, to show off the technology's progress. Experts in the industry commented AI's potential to enhance stories without taking over the characters or plots.

Key details:

  • Over 3,000 film submissions, up from 300 last year

  • 'Get Me Out' won the Grand Prix, using tools like Luma AI, Runway, and ComfyUI

  • Veteran producer Joel Kuwahara wants to use AI to complement stories

  • Director Paul Trillo says major filmmakers adopting AI will further its acceptance

Why it matters: The festival showed how generative AI has moved from experimental novelty to a valuable filmmaking tool to improve stories and visuals. With more filmmakers using it, Hollywood's is likely to increasingly rely on AI for movie storytelling.


Quick news


🥇 New tools

That’s all for today!

If you liked this newsletter, share it with your friends and colleagues by sending them your invite link : {{rp_refer_url}}.