What is today’s beat?

RELEASES
🧨 Anthropic Economic Index to AI impact in labor markets
🧨 An overly bias look at this week in AI

NEWSROOM
🗞️ Saudi Arabia Invests $1.5B in AI Expansion with Groq
🗞️ Google Enhances Urban Mobility with Scalable Digital Twins
🗞️ Musk and his $97B Bid to Acquire OpenAI
🗞️ ROOST Initiative Launches Open-Source AI Safety Tools

BUILDER BYTES
⭐️ Developing Agentic AI with DeepSeek-R1 and Amazon SageMaker
⭐️ Saving 77% of the Parameters in Large Language Models
⭐️ Code and paper for s1, simple test-time scaling

COMMUNITY
🤩 What if all the world's biggest problems have the same solution?
🤩 In the short term, more chaos”: What’s next for API design
🤩 AI In the Classroom with Irina Jurenka

FOLLOW US ON SOCIAL FOR MORE

Your FREE newsletter
 share or subscribe 
to show support

🧨 Anthropic 🧨

ECONOMIC INDEX,
TRACKING IMPACT ON LABOR MARKETS

In the coming years, AI systems will have a major impact on the ways people work. The Anthropic Economic Index is an initiative aimed at understanding AI's effects on labor markets and the economy over time, so that others can watch it. It’s kind of like the World Economic Forum’s report.

• Concentration in Technical Tasks: AI usage is predominantly observed in software development and technical writing, with these areas accounting for nearly half of all AI-related activities.

• Augmentation vs. Automation: The data indicates that 57% of AI applications are augmentative—assisting and enhancing human work—while 43% are automative, where AI performs tasks independently.

• Occupational Reach: Approximately 36% of occupations incorporate AI in at least a quarter of their tasks, whereas about 4% utilize AI in over 75% of their tasks.

• Wage Correlation: AI usage is more prevalent in mid-to-high wage professions, such as computer programming and data science, and less common in both the lowest- and highest-paid roles.

Read more about the economic Index.

🧨 RELEASES 🧨

What happened this week?

While the buzz of the 2025 AI Summit in Paris is echoing across the world, which is huge for policy makers and startups and global corporations, we’ve seen some action on the technology side that is doing the same. DeepSeek, OpenAI, and Mistral AI take the cake this week!

• DeepSeek's Janus-Pro 
  An open-source multimodal AI model available in 1B and 7B parameter versions, integrates a unified transformer architecture for both understanding and generation tasks AND supports native 1024x1024 image generation with an average inference time of 2.4 seconds. 84.2% accuracy beats openAI’s DALE-E 3 by a wopping 0.7%.

• OpenAI's o3-mini
  Delivers o1-level performance with faster processing. Go to OpenAI to find out more, and just because we use it on a daily basis doesn’t mean we have to like Sam. That’s all for now.

• Mistral AI's Small 3
  Same performance as Qwen 32B, Llama 3.3 70B, and OpenAI-GPT-4o mini, but its faster (150 tokens per second throughput, 11-12ms latency per token). And, you can get it on HuggingFace.

These developments highlight another week in AI. The open-source nature of Janus-Pro and Small 3, along with their flexible licensing, encourages me to take automated generated sentences and ingest personal flavor and humor. Collectively, we’ve successfully made it past week 7 of the year without AGI taking over. But, the hype man isn’t done yet, there is so much more in store!

On a very positive note, Le Chat by Mistral AI, a French model provider, surpassed ChatGPT as most downloaded app in France. If you are European, you are kind of obliged to check it out: Le Chat.
 

🗞️ NEWSROOM 🗞️ 

What’s hot in tech right now?

Saudi Arabia Invests $1.5B in AI Expansion with Groq

Saudi Arabia commits $1.5 billion to enhance its AI infrastructure, partnering with Groq to deliver advanced AI inference capabilities and launching GroqCloud™ services from a data center in Dammam.

Google Enhances Urban Mobility with Scalable Digital Twins

Google introduces a scalable, high-resolution traffic simulator to improve urban mobility, utilizing efficient algorithms to optimize travel demand estimation and traffic management.

Musk and his $97B Bid to Acquire OpenAI

A consortium led by Elon Musk offers $97.4 billion to acquire OpenAI, aiming to revert it to a nonprofit research lab, though the bid faces rejection from OpenAI's CEO.

ROOST Initiative Launches Open-Source AI Safety Tools

At the Paris AI Summit, the ROOST initiative debuts, aiming to enhance AI safety with open-source tools, emphasizing scalable, interoperable safety infrastructure for the AI era.

⭐️ BUILDER BYTES ⭐️

What’s hot for builders right now?

Developing Agentic AI with DeepSeek-R1 and Amazon SageMaker

Explores building autonomous AI agents using DeepSeek-R1 models, CrewAI framework, and Amazon SageMaker AI to enhance decision-making and efficiency in complex tasks.

Saving 77% of the Parameters in Large Language Models

A Github repo based on a paper working on phi-3-mini-4k-instruct to reduce 77% of its parameters while still achieving SOTA results.

s1: Simple test-time scaling
Paper with a Github repo, Minimal recipe for test-time scaling and strong reasoning performance matching o1-preview with just 1,000 examples & budget forcing.

🤩 COMMUNITY 🤩

What’s the latest beat?

THANK YOU

Looking to promote your company, product, service, or event?