Reading Sara Hooker’s On the Slow Death of Scaling: bigger models + more compute ≠ guaranteed better performance anymore.

Progress is increasingly driven by data, optimization, architecture, and inference-time strategies.

Real-world data imperfections are inevitable. We gotta build a ML system that can handle these imperfections, not just in ideal conditions aka lab settings

Data naturally degrades over time due to inconsistency, missing values, and evolving business conditions

📖Data drift - when the data is changing (or drifting) over time. Historical data no longer represents what is currently happening

David Silver! #deeplearning #deepmind youtu.be/zzXyPGEtseI?...

Is there a similar #datavisualization for #NYC?

www.reddit.com/r/MapPorn/s/...

🌍📊 A one-stop shop for macroeconomic data-track inflation, interest rates, GDP, and more across global markets. Whether you’re analyzing trends or building models, this dataset offers key insights into economies.

📖 Link: 👉 www.globalmacrodata.com/index.html

#Macroeconomics #Finance #MarketTrends

Many people have asked me about the France Action Summit.

I think a summit is typically most valuable as a catalyst, not as a solution in itself.

But, will share some observations.

“by predicting representations at a high level of abstraction rather than predicting pixel values directly, the hope is to learn directly useful representations that also avoid the limitations of generative approaches, which underlie the large language models”

#i-jepa #ai @yann-lecun.bsky.social

Where does black fall on the color spectrum? A color scientist explains

🔥 The future of AI is OPEN! 🔥
Open source AI tools are driving the future.

📖 Dive deeper into the LF Research Shaping the future of GENAI report!
➡https://www.linuxfoundation.org/research/gen-ai-2024

#AI #OpenSource #MachineLearning

Woah… there is a life outside of numbers

Plotted my velocity during a bike route in #NYC using the #clusterduckprotocol

#iot #opensource #lora #datavisualization

Open source you say 💙