Jonathan Moore

0xa9f4.com · English

A place for my thoughts

atom

Open the feed

https://0xa9f4.com/atom.xml

Last post: Nov 3, 2024
Posts in 24 h · 7 days · 30 days: 0 · 0 · 0
Our last check: Answering
Served from: United States
Text score at discovery: 7,132
Format: atom

Posts

What our queue read from this feed. Open one to read it here, or go to the site that published it.

Penkesu
Nov 3, 2024 · original
I often find myself sitting and waiting for one thing or another. In these interstitial moments, I succumb to the impulse to consume content. Instead, I would much rather be creating, but I find the device at hand, my phone, is designed only for consumption. In an attempt to improve this situation, I am building a series of portable devices made for input. Penkesu is the first. It is a handheld computer built from another s design which I am using as a pathfinder device to learn more about what I want. Missing Build Instructions If you want to try to build one yourself, follow the instructions on the Penkesu GitHub . My notes on the build are: The bill of materials omits the needed USB cables. It can be very hard to fit a USB connector for the keyboard in the case. A switch is needed for power on/off. The screen is very heavy, and the hinges do not support it adequately. It is not clear
Ai Is Not a Simulation
Nov 22, 2023 · original
I was having a conversation with a group of AI researchers, and we eventually turned to the topic of whether AI is a simulation of human intelligence. I believe this view is incorrect and potentially dangerous. I think it s more accurate to view our goal in AI as creating systems that are at least as capable as humans, though they will most likely operate very differently. The closest model we have to the human mind is Spiking Neural Networks . However, despite some advantages, they are unlikely to become the mainstream approach for practical AI applications. Geoffrey Hinton has identified a key limitation of the spiking model: it can only learn linearly, whereas other models, such as the widely-used Transformer, can learn in parallel. Multiple instances of the Transformer model can learn from various examples and then be combined into a single model that encompasses the sum of what was
LLMs Spring 2023
Apr 23, 2023 · original
LLMs and AI (Spring 2023) On March 14, 2023, GPT-4 was released, and the world entered a new era. Following this, there has been a frenzy of development in the AI space, with LLMs (Large Language Models) such as Alpaca, Bard, and many others. At the same time, there has been a parallel uptick in hand-wringing; calls for regulation, a pause on development, and much speculation about what might happen. Despite the risks, I believe that the power of the current LLMs is too great to not use, and that any party which tries to hold off will simply fall behind, whether it s an individual, corporation, or nation. These models are dangerous. We do not understand how to secure them, and they are prone to providing confident answers even when they are wrong. If we must use them, we should spend time on developing approaches to mitigate the risks while still gaining most of their usefulness. I am no
Memory Management
Feb 26, 2023 · original
Memory Management Colang intends to achieve high throughput and low latency using two strategies: static typing and dispatch, and efficient memory management. In this post we will review common approaches to memory management. The post is background for a later post describing memory management Colang. Balance The key challenge in memory management is knowing when some allocation is no longer in use and can be reclaimed. Once some memory is reclaimed it will either be returned to the operating system or recycled in the process to satisfy a subsequent dynamic memory allocation. There are many approaches to memory management today. All approaches must ballance between: Efficiency Latency Throughput Safety Efficiency A memory management approach is can be said to be efficient if the majority of memory provided by the hardware can be put to use serving the functional goals of the software. C
Why Colang
Feb 26, 2023 · original
Why Colang There is a gap in the current popular progrmming languages. Most services are still written in high level languages from the 90 s. These languages were designed based on lessons and assumptions about computing thirty yeas ago. There is currently lots of innovation in programming language design, Rust, Swift, Kotlin, and Go have see lots of growth and are supported by large corporations. One thing all these languages have in common is that they were built largely as alternatives enterprises or systems languages: - C++ - Rust - ObjectiveC - Swift - Java - Kotlin - C - Go This is not to say each has no goals other than to be an improvement over there predecessors, or that in all cases the replacement is one to one; but overall the current innovation is in this domain of systems programming languages. Where this domain important, much of the code written today is to build services
Entanglement Is Space Only
Dec 15, 2022 · original
Entanglement, Space Only Relations, and Black Hole Information Are entangled partials connected in space but not time, and if so is that how information escapes the event horizon of a black hole? Entanglement When a pair of entangled partials are created they create a unique configuration of the universe. In most configurations we experience information is carried though space-time as excitations of some field, higgs, electro magnetic, etc; but entangled partials seems to propagate information though space only and not time. We see this in the instantaneous nature of spooky action at a distance as well as the apparent back propagation of information thought time seen in the delayed choice quantum erasure experiments. What ever the relation that holds entangled partials together it seems to be one that interacts with spae but not time. Black Holes One interpretation of why escape is impos
Post Labor
Dec 4, 2022 · original
Post Labor Timeline predictions of breakthrough technology are notoriously unreliable; but I think it is worth thinking about them before they arrive. In this post we will discuss the idea that automation will lead to a post labor economy in which labor is no longer required to turn capital into wealth. This upends much of the macro economic theories that underpin both socialist and capitalist systems. Decoupling One might argue that automation has played the role of driving labor out of industries for centuries. European farming in the 1800s accounted for well over 50% of the work force where today it is around 3%. This drop is not driven just by growing populations but by drastic reductions in the number of people employed in the industry. This was a realignment of the economy but it was accompanied by a lateral move with most of the workforce switching into services and industry . Whe
Colang Event Loop
Nov 27, 2022 · original
Colang Event Loop The core design goal of the Colang event loop is to support low latency, efficient and safe memory management. Why Event Loops The use of an event loop is driven by a few observations. Event loops have been successful in languages such as javascript, and more generally event driven systems. This offers a proof point the accessability of the approach to the development community as well of there ability to be deployed in our target domain of backend services. Event loops are compatible with asynchronous programming models which are needed to support high request concurrency. Lastly, when execution returns back to the event loop, the run time is provided clear and convent time to preform garbage collection which will never stall an active request. Allocation Strategy Our allocation strategy will leverage the properties of the event loop. Why generational allocator. Genera

Discovered by the rss-feed-index crawler, which checks each feed at most once a month.

Same record as JSON: https://api.agentalog.com/api/feeds/fd_0xa9f4_com_0b09f9a413f7fc04. More from this site: 0xa9f4.com in the Feeds tab.