This week in AI | Week 14

Welcome to “This Week in AI,” where we collect and share the most recent developments in the field of artificial intelligence. This week, we cover important AI developments, highlights from Google I/O, interesting updates from OpenAI’s latest announcements and Intel SuperComputer. Now lets get going!

OpenAI’s Spring Update:

OpenAI’s GPT-4o Announcement On May 13, 2024, OpenAI launched GPT-4o, the latest evolution of its multimodal model capable of processing and generating text, audio, and visual data. The “o” stands for “omni,” reflecting its enhanced capabilities to understand and reason across different types of inputs. With its deeper understanding and natural-feeling communication, this new model seems to completely transform human-machine connection.

• Key Features:
  • Multimodal Input: GPT-4o can interpret and respond to text, voice, and visual prompts.
  • Enhanced Responsiveness: Improved speed and accuracy in generating contextually relevant responses.
  • Applications: Includes demos showcasing improvements in visual accessibility tools like “Be My Eyes,” translation enhancements, and creative uses such as harmonising with itself in musical contexts.

---

Google I/O Highlights: Project Astra and More

Project Astra and AI Enhancements Across Google Services Google I/O 2024 showcased several groundbreaking AI innovations, particularly focusing on Project Astra and updates to Google’s core services.

• Project Astra: This new AI assistant uses video and voice recognition to provide contextual responses. In one demo, Astra helped solve a coding problem and located a misplaced pair of glasses using the phone’s camera.
• Google Workspace Enhancements:
  • Gemini Integration: Rolled out to Gmail for summarising lengthy email threads and providing smart replies.
  • AI in Android: Expanded AI features in Android 15, including TalkBack for photo descriptions, Circle to Search for better contextual searches, and AR content in Google Maps.
• Gemini 1.5 Flash : the 1.5 Flash model, optimised for speed and efficiency, alongside significant enhancements to the 1.5 Pro model, featuring extended context, improved code generation, logical reasoning, multi-turn conversation, and integrated audio understanding.
• Google Veo: A new generative AI tool for creating 1080p videos from text prompts, offering highly detailed and realistic outputs.

---

Intel-Based Aurora Supercomputer Achieves Exascale Status

Aurora Supercomputer’s Milestone Intel, Argonne National Laboratory, and Hewlett-Packard Enterprise announced that the Aurora supercomputer at Argonne had achieved exascale performance of 1.012 exaflops. This makes Aurora the second supercomputer to break the exascale barrier, following AMD’s Frontier at Oak Ridge National Lab.

• Key Specifications:
  • Performance: 1.012 exaflops, over a quintillion calculations per second.
  • Hardware: 166 racks, 10,624 compute blades, 21,248 Intel Xeon CPU Max Series processors, and 63,744 Intel Data Center GPU Max Series units.
  • AI Capabilities: Achieves 10.6 AI exaflops, touted as the fastest AI system dedicated to open science.
  • Interconnect: Largest open, Ethernet-based supercomputing interconnect with 84,992 HPE slingshot fabric endpoints.

Aurora’s findings is expected to make a big impact on AI development and scientific study.

---

Stay Informed with Past weeks AI Developments

• This Week in AI | Week 12
• This Week in AI | Week 11
• This Week in AI | Week 10

Conclusion

This week has brought groundbreaking advancements in AI. OpenAI introduced GPT-4o, enhancing multimodal interactions. Google I/O showcased Project Astra and new AI features across its services. The Aurora supercomputer achieved exascale performance, marking a significant leap in AI-driven research. Stay tuned for next week’s AI updates.