#header-inner img {width: 900px; margin: 0 auto;} #header-inner {text-align: center;}

Future of AI

News on Artificial Intelligence in Education and Libraries

Saturday, February 1, 2025

THIS WEEK: in AI Products (February 1, 2025)

 

OpenAI 11415

  • Operator Agent Launch: Introduced a general-purpose AI agent capable of autonomous web browsing to handle tasks like travel bookings and restaurant reservations (currently in research preview for $200/month Pro plan users)
  • o3 Model Family: Released successors to its reasoning-focused o1 models with improved multimodal capabilities, though availability remains limited to safety researchers
  • ChatGPT Accessibility:
    • Launched 1-800-CHATGPT phone service with 15 free minutes for U.S. callers
    • Enabled phone number-only signups in U.S./India beta (email required for paid upgrades)
  • Content Partnerships:
    • Finalizing Media Manager tool for creator opt-outs from AI training by 2025
    • Expanded Stack Overflow integration with code attribution features
  • Policy Developments: Published new NSFW guidelines exploring explicit content allowances while facing criticism over ethical implementation

Microsoft 281819

  • Copilot Evolution:
    • Added real-time voice interaction and screen context understanding
    • Launched "Daily" feature suggesting AI-powered actions based on user activity patterns
    • Introduced Vision capabilities for image/webpage analysis
  • Windows 11 Integration:
    • Rolling out Copilot+ PC optimizations for AI workloads
    • Testing WIP features like automated meeting summaries and smart file organization
  • Healthcare Focus: Released specialized medical AI models in Azure Studio for clinical data synthesis and nurse documentation assistance

Google 31320

  • Gemini Ecosystem Expansion:
    • Launched Gemini 2.0 with improved reasoning capabilities
    • Previewed AR glasses powered by Gemini visual processing
    • Introduced Android XR platform for spatial computing
  • AlphaChip Breakthrough: Deployed reinforcement learning AI that designs computer chips 100x faster than human engineers (already used in 3 TPU generations)
  • Search Overhaul: Testing AI-generated inspirational content for travel, dining and shopping queries

Anthropic 4814

  • Agent Framework: Released computer-controlling AI with 14.9% task success rate in early trials, using human-like UI interaction
  • Cloud Partnerships:
    • Deepened AWS collaboration for Trainium chip development
    • Preparing Claude web search functionality to rival ChatGPT Browse

Perplexity 5915

  • Pro Search Upgrade: Enhanced research capabilities with multi-step query planning and synthesis
  • DeepSeek Integration: Offered privacy-focused access to controversial Chinese model through $20/month Pro tier
  • Strategic Moves: Submitted revised TikTok merger proposal allowing 50% U.S. government ownership stake

Grok (xAI) 61016

  • Free Tier Launch: Opened access to all X users with:
    • 3x faster response speeds
    • New multilingual support
    • Tweet analysis via timeline-integrated Grok button
  • Upcoming Features:
    • Standalone app development
    • Image generation API expansion
    • Persistent conversation memory improvements

Other Notable Updates

  • DeepSeek: Open-source model sparked global debate over Chinese data practices 911
  • Alibaba: Released Qwen 2.5 claiming parity with GPT-4 and Claude 3 7
  • Meta: Faced backlash over AI training data practices while advancing AR glasses 20
This week saw intensified competition in autonomous agents, with OpenAI's Operator and Anthropic's computer-control systems pushing toward practical task automation. Accessibility improvements dominated consumer offerings, from free Grok access to ChatGPT's phone service, while ethical concerns persisted around content sourcing and international data flows.