Winners and Waste, Federated Learning, Robotic Gardening
The performance gap between the equal-weight S&P 500 and the regular-weight S&P 500 is on track for it’s largest gap since the metric started 30+ years ago. This indicates a growing split between haves and have-nots. The winners are obvious, they are the technology leaders with the largest balance sheets like NVidia, Meta, Google, Apple, and so on. It also means the have-nots are not advancing, and there are many have-nots. The Nasdaq is up 30% YTD while 2/3rds of US companies have YTD losses.
On January 4th, 2022, Amazon had the largest single-day gain with $190B (+12%). NVidia now holds the second spot with $184B on May 25th, 2023. NVidia’s single day gain is directly tied to the feverish spending to expand AI-related computing power. Three years ago analysts expected NVidia to reach $30B in revenue by 2027. Today analysts expect $83B in revenue by 2027. NVDA achieved its $1T market cap with just 27,000 employees, less than 15% of each of the other $1T market cap employee bases and less than 2% of the Amazon employee base.
The US has a spending problem. If you’ve ever had a family member or friend who can’t control their spending, then you know exactly what I mean. Improper payments (spending errors) were $247B in 2022. What about $16 muffins and $8 coffee? Or $3.6m to build a parking lot that was poorly built and required $500k to tear down. Or the LA-SF high-speed rail, which started in 2008 and was supposed to be finished by 2020 with a budget that has grown from $33B to $105B. Or the 3.5 mile $11B subway in NYC, the most expensive track in the world. Or the famous NYC 2nd Avenue subway, which has allocated more to consultants than actual construction.
You can’t force people to change, but remarkable change is possible. Our Social Security system is set to run out of money by 2034. However, minor changes can improve the math dramatically, like removing the COLA (cost of living adjustments), increasing the benefits age to 67, (or maybe even 70?), means testing, increasing the contribution level and cutting the distributions slightly. Unsustainable math never works out and you can’t wish it away. The US government and leadership must shift its mindset as it risks wasting an embarrassment of riches. Waste not, want not.
Federated learning protects user privacy while continuously updating and training machine learning models. It’s a valuable exercise to keep up with the new lingo as AI models grow.
With federated learning, user data stays on their machine. The ML model is sent to the user’s device rather than their data sent out to a centralized database. After the model updates on the local machine, it’s sent back to the centralized servers. This removes the need to store sensitive or personalized information in a centralized location. With privacy concerns at the forefront, this term should help people understand how their data can remain secure while still being used to train and improve large machine-learning models.
Researchers at UC Berkeley were able to train a robot (called AlphaGarden) to tend to a polyculture garden. Polyculture refers to the 32 different plant species that share a small space. The robot kept up with the humans while using 44% less water. For now, the robots are more expensive than human support, but that will change.
John Deere recently held its annual tech summit where it announced an intention to “make sure every one of the 10 trillion corn and soybean seeds can be planted, cared for, and harvested autonomously.”
If you consider how AI will play a role in controlling and training robots just in agriculture alone, it’s easy to see how robotics and machine learning will have a massive economic and societal impact in the coming decade.
Weekly Articles by Osbon Capital Management:
"*" indicates required fields