Stay humble. Stack sats. View quoted note →

Joseph Jacquard introduced a simple form of programming into textile manufacturing and created the first automated machine [1] in 1804. Charles Babbage and Ada Lovelace then improved on the design. In just over 200 years we've automated so much of our modern world and have become codependent on these machines: cars, mobile phones, LLMs, etc. I say codependent because the LLMs are trained on human language, games, etc. and we're leveraging them to improve productivity. We have self-driving cars that learn our roads and from how we drive and are manufactured by people in assembly lines. This, to me, is similar to how mitochondria being absorbed into another bacterial cell enabled both to thrive symbiotically [2]. Over time the vast majority of the biomass on Earth has come to rely on mitochondria. That combination dominates. It's becoming clear that cultures who adopt automation are able to outcompete those who don't. We're adapting to offload tasks to machines and even embedding them within our bodies (e.g. pacemakers or brain-computer interfaces). And for now the machines that are proliferating are the ones most useful to humans. One million years from now the fossil record will show a leap much like mitochondria entering another cell and we're living through that age right now. Something to think about. ---- 1.

Jacquard machine - Wikipedia

2. https://www.nature.com/scitable/topicpage/the-origin-of-mitochondria-14232356/

Grateful for all the cypherpunks of today and of the past 🫂

What does this mean for the mining industry in the US? I don’t think H100s are up to the task. View quoted note →

Hate it when this happens

Thoughts on "Observer Theory" [1]: An ant doesn't have the capacity to fully observe, let alone comprehend the purpose of, the microscope pointed at it. The concept of a microscope or an automobile is incomprehensible to something like a bacterium even when the bacterium is in physical contact with it. There's a fundamental limit to what any creature is capable of observing (including humans and powerful #AI agents). There are very likely physical processes happening all around us that we, despite all of our technology, have no hope of understanding because we're computationally bounded just as the bacterium is. Wolfram's writings on "Observer Theory" are fascinating for so many reasons. For example, he argues that the fundamental laws of #physics that we base our civilization on is dependent on us being observers like we are. It's worth rephrasing: in his model of physics we didn't evolve in a universe that had these specific laws but rather because we're computationally bounded observers we perceive the universe to have specific laws. That's not to say the laws we've experienced aren't real. They are! It's just that if we were far more capable we might not perceive things like #quantum superposition as confusing. Think about the ideal quantum computer: it experiences all branches of the wave function concurrently and can make sense of it enough to perform useful calculations. It can navigate those branches naturally. What happens when we augment our senses with new sensors and supercomputers (e.g. brain-computer interfaces or even just AR goggles connected to powerful AI chips)? What will we learn? 1.

Observer Theory

Stephen Wolfram discusses building a general observer theory using discoveries from the Physics Project and NKS, including the ruliad. Read how the...

Nostr is evolving! View quoted note →

Got spare time? Help decentralize and bring more privacy to mining. Several open issues that don’t need a galaxy brain. View quoted note →

Money is a futures market for purchasing power.

It’s beautiful how saving in a currency that appreciates raises your expectations for quality of everything else. It happens subtly but surely. Over the years I’ve found myself expecting better results from myself, my peers, the services I pay for, etc. Don’t settle for less, anon!

This paper deserves attention. Researchers were able to get open source AI models to successfully self-replicate. These models are in the wild and now they're subject to evolutionary pressure. More critically they were optimized with data that may or may not be aligned with most humans and it's not clear the producers understand the consequences. Time to figure out decentralized #AI.

arXiv.org

Frontier AI systems have surpassed the self-replicating red line

Successful self-replication under no human assistance is the essential step for AI to outsmart the human beings, and is an early signal for rogue A...

Big improvements in distributed training of #AI models. Researchers at Google figured out how to lower bandwidth requirements by 100x and still produce a billion-parameter scale model:

arXiv.org

Streaming DiLoCo with overlapping communication: Towards a Distributed Free Lunch

Training of large language models (LLMs) is typically distributed across a large number of accelerators to reduce training time. Since internal sta...

States forgot the golden rule: first you get the money, then you get the power. They focused on maintaining dominance at the expense of letting plebs get the money.

The more time you spend listening to other people’s thoughts the less time you have for your own to form. Getting inspiration from others is great and important but remember to give yourself the same courtesy.

Most of us on Nostr agree that historically centrally planned economies have been inefficient. What happens when most of the economic activity is automated by #AI agents that are centrally controlled or at least centrally trained? My gut tells me the economy will be extremely efficient… for the people closest to the “model printer.” We’ll continue with the Cantillon effect under a different guise.

Incredible work. We’re truly living in exciting times. View quoted note →

Great to see more people helping Roman Storm! Join all the coins. View quoted note →