Must See AI News This Week - NVIDIA DIGITS, OpenAI O3, Microsoft Phi-4, ElevenLabs Flash Speech Model
OpenAI debuts O3 reasoning model with great math skills, NVIDIA launches $3K personal AI supercomputer at CES, Microsoft open-sources Phi-4 & invests $3B in India, ElevenLabs drops Flash Speed Model
👉 Most Interesting This Week
1️⃣ OpenAI's O3 Reasoning Model Debuts
OpenAI announced O3, a new AI model focused on advanced reasoning capabilities
Key performance metrics:
87.7% accuracy on graduate-level science questions
25% score on FrontierMath (previous models under 2%)
2,727 ELO rating in competitive programming
Uses "verifier" system to evaluate hundreds of potential solutions
Development stats:
Cost: $350K
Processing time: 16 hours
Public access timeline:
O3-mini launches late January 2025
Full O3 release planned for Q1 2025
2️⃣ NVIDIA Unveils Project DIGITS at CES 2025
Launched $3,000 personal AI supercomputer powered by GB10 Grace Blackwell Superchip
Delivers 1 petaflop of AI performance
Can run 200B parameter models on standard power outlet
Features new RTX 50 series GPUs with Blackwell architecture
Introduces Cosmos platform for physical AI:
Trained on 20M hours of video data
Enables physics-aware simulations
Used by companies like Uber for autonomous systems
3️⃣ Microsoft Open Sources Phi-4
Released 14B parameter model on Hugging Face under MIT license
Outperforms larger models in math and multitask language understanding
Features 16K token context window
Trained on approximately 10 trillion tokens
Uses innovative synthetic data generation techniques
📱AI Business Moves
1️⃣ Microsoft's $3B India AI Investment
Announced $3B investment over next two years for AI and cloud infrastructure
Plans to establish new data centers
Goals include:
Expanding Azure cloud capabilities
Training 10M Indians in AI skills by 2030
Part of larger $80B global AI infrastructure investment plan
2️⃣ Anthropic Funding Round
Nearing completion of $2B funding round
Projected valuation: $60B
Follows significant model improvements and enterprise partnerships
Strengthens position against OpenAI and other competitors
👇 AI in Action
1️⃣ ElevenLabs Flash Speech Model
New low-latency speech generation model
Features:
75ms generation time
Available in two versions: Flash v2 (English) and Flash v2.5 (32 languages)
Optimized for real-time conversational AI
Outperforms competitors in blind tests
Accessible via API and Conversational AI platform