AI News Breakdown

OpenAI Announcements

• O3 Mini Release

  • Released last Friday, outperforming most models except for O1 Pro.
  • Excels in math, coding, and software engineering.
  • Available in all user tiers and API access.
  • Pro users: Unlimited access; Plus and Team users: Triple rate limits.
  • Free users can try O3 Mini in ChatGPT by selecting "reason".
  • O3 Mini can be combined with search even on free plans.
  • Chain of thought feature updated for both free and paid users on February 6.
  • Criticism on summarized chain of thought for debugging purposes.

• Deep Research Release

  • Available for $200/month Pro users.
  • Provides in-depth analysis and strategy formulation, e.g., for YouTube channels.
  • Results in detailed reports and actionable plans.
  • Combines research and reasoning for comprehensive outputs.
  • Rolled out to all Pro users globally.

• Humanity's Last Exam Benchmark

  • New models tested; O3 Mini High at 13%, Deep Research at 26.6% accuracy.

• Additional OpenAI Updates

  • ChatGPT search available without sign-up.
  • Memory limit increased by 25% for Plus, Pro, and Team users.

Google Announcements

• Gemini 2.0 Release

  • Includes Gemini 2.0 Flash, Flash Lite, and Pro models.
  • Flash models have 1 million token context window; Pro has 2 million.
  • Competitive pricing for developers.
  • Free access to Gemini models at AISTudio.Google.com.

• Image Generation

  • Imagine 3 AI image generator available via API.
  • Imagine 3 ranked top in text-to-image leaderboard.

Other AI News

• Mistral AI

  • New version of Le Chat released.
  • Offers fast processing at 1000 tokens/second.

• Anthropic's Clod

  • Lyft using Clod for customer service, reducing resolution time by 87%.

• Amazon Alexa Event

  • Anticipated integration with Anthropix Cloud.

• GitHub Copilot Updates

  • New agent mode with self-healing capabilities.

• Cursor Growth

  • Fastest growing SaaS company, reaching $100 million ARR in one year.

AI in Creative Applications

• Grok on X

  • New image editing capabilities.

• Pika Labs Updates

  • Pika Scenes for AI-generated pet videos.
  • Pika Ditions for integrating images into videos.

• Topaz Labs

  • Project Starlight for video upscaling and restoration.

Research Highlights

• OmniHuman1

  • Combines image and audio to create deep fakes.

• VideoJam

  • Improved AI training for coherent video physics.

Miscellaneous

• Legislative News

  • Proposed bill to make downloading Deep Seek illegal.

• Music and AI

  • Beatles win Grammy for AI-assisted song.

Conclusion

• Recap of a busy week in AI with ongoing developments.
• Encouragement to subscribe for regular AI updates.
• Future tools and newsletter available at futuretools.io.