AI and Tech Updates

Overview

• Busy week for AI announcements while presenter was at Disneyland.
• Two major events: OpenAI release and Apple iPhone event.

OpenAI 01 Preview Model

• Naming and Release:
  • Originally teased as Qstar, then Strawberry.
  • Current name: OpenAI 01, signaling a new naming scheme for models.
  • Future models expected to follow: OpenAI 02, 03, etc.
• Access and Usage:
  • Available for pro and enterprise plan users.
  • Defaults to GPT 4.0; new models accessible via a dropdown menu.
• Model Features:
  • Recommended for advanced math, logic, and reasoning.
  • Implements "chain of thought" prompting for thoughtful responses.
  • Slower response times due to thorough thinking process.
  • Example use: Solving complex puzzles and math problems.
• Model Performance:
  • Ranks in the 89th percentile in competitive programming.
  • Outperforms GPT-4.0 in several benchmarks like math and coding.
  • Features a Mini version, more cost-effective and faster, but only for paid members.
  • Criticism exists around pricing and slower response time.
  • Future potential for faster model releases due to reduced pre-training time.

Apple iPhone Event

• Focus:
  • Update on iPhone 16, Apple Watch, and AirPods.
• AI Features:
  • Most had been previously announced at WWDC.
  • Features include email summarization, photo cleanup, AI art generation, and more.
  • AI translation added to Apple Watch.
  • New interaction methods with Siri via nodding or shaking head.
  • Private cloud compute for larger models.
  • Announced features will roll out over time, not available immediately.

Adobe's Firefly Text-to-Video Model

• Overview:
  • Generates short, high-quality videos based on text prompts.
  • Claimed to be ethically sourced and commercially safe.
  • Examples include various creative and realistic animations.

Other AI and Tech Updates

• Mistral's Pixtrel 12b:
  • First open-source model to accept images as input.
• Google Notebook LM: Audio Overviews:
  • Converts document discussions into podcast-style audio.
• Amazon Audible AI Narration:
  • AI voice clones for narrators to speed up audiobook production.
• Suno Covers Feature:
  • Transforms music or voice clips into different styles.
• Meta's AI Content Labels:
  • Less prominent labeling of AI-generated content.
• Roblox and Cybever:
  • AI tools for creating detailed 3D worlds.
• Meshi Version 4:
  • Enhanced 3D object generation from text prompts.
• PlayStation 5 Pro:
  • Uses AI to upscale graphics, but criticized for lacking a disc drive.
• DeepMind Robotics:
  • New robot capabilities include shoe-tying and clothing handling.

Conclusion

• Presenter apologizes for the delay in video release due to Disneyland trip.
• FutureTools.io recommended for daily AI news updates.