Your Own AI Stock Trader

"Robinhood launched agentic trading and an agentic credit card today that will allow AI agents to trade equities and make credit card purchases on customers' behalf."

Costco Theory Of The Internet

"The Costco theory of the internet"

Using AI To Sell Your House

NYT: "I Tried to Sell My House With a Chatbot" 

In the end, using A.I. netted me more than $90,000. That includes the premium over the asking price, plus the roughly $36,000 in fees I didn’t pay. I'm not sure how replicable my experience would be for anyone less versed than me in technology.

Less Drag

"A Fundamental Principle of Aeronautical Engineering Has Been Overturned"

Aiko Yakino, associate professor at Tohoku University's Institute of Fluid Science, and her research group were the first in the world to demonstrate that aerodynamic drag can be reduced by up to 43.6 percent simply by applying distributed micro-roughness (DMR), a surface roughness so fine and irregular that it cannot be distinguished by the naked eye. 

This technology is fundamentally different from the rivulet (“shark skin”) process, which is a known air-drag-reduction technology. The rivulet process mimics the fine longitudinal grooves in shark skin, and by carving grooves approximately 0.1 millimeter wide along the direction of airflow, it aligns the vortices that occur near the wall surface of turbulent airflow areas. DMR, on the other hand, delays the switch from laminar to turbulent flow by means of random and minute irregularities. The flow zones it affects and the mechanisms it employs are based on completely different concepts.

Our Latest Overlords

"Scientists Build a Living AI Device Using Real Brain Cells". (Via H.R.)

Do LLMs Need Sleep?

Maybe you don't want your LLMs to work 24/7: "Do Language Models Need Sleep? Offline Recurrence for Improved Online Inference".  Abstract: 

Transformer-based large language models are increasingly used for long-horizon tasks; however, their attention mechanism scales poorly with context length. To handle this, we study a sleep-like consolidation mechanism in which a model periodically converts recent context into persistent fast weights before clearing its key-value cache. During sleep, the model performs offline recurrent passes over the accumulated context and updates the fast weights in its state-space model (SSM) blocks through a learned local rule. During inference, this shifts extra computation to sleep while preserving the latency of wake-time prediction. We test our method on controlled synthetic tasks, including cellular automata and multi-hop graph retrieval, as well as a realistic math reasoning task, on which a regular transformer as well as SSM-attention hybrid models fail. We then show that increasing sleep duration for our models improves performance, with the largest gains on examples that require deeper reasoning. 

(Via S.S.)