🌸 Fri, Aug 2, 2024 🌸
Astrophysicists say the age of the universe is 13.7 billion years, but do we agree on what is the duration of a year? 365, 365.25, or 365.24219?
🌸 Thu, Aug 1, 2024 🌸
About 3 weeks ago, Physics Girl made an 11-hour live stream to raise awareness about long-COVID:
https://www.youtube.com/live/v8HWt9g4L0k?si=cs7IWtIE4pjvvlOm
I tried to watch the stream multiple times, as I’m interested due to my own long-COVID situation, but I only managed to watch some pieces of the interview and couldn’t watch the whole stream. I tried different LLMs to give me a summary of the stream, and finally, after multiple hours of “torturing” my laptop, Llama 3.1 (70B parameters) provided some summarization.
Disclaimer: Large language models may give incorrect results. Take the summarization with a grain of salt, do your own research, and watch the stream.
Summary:
The live stream was a 12-hour fundraiser for the Open Medicine Foundation, which aims to support research and clinical trials for ME/CFS and long COVID. Diana, a physics girl, has been bedridden for over a year due to ME/CFS and long COVID. The stream featured interviews with experts, including Dr. David Kaufman, Dr. Raven Baxter, Linda Tannenbaum, and Dr. David Patrino. They discussed various aspects of ME/CFS and long COVID, including the importance of pacing, the role of exercise, and the need for more research.
Therapies that appeared to be useful:
- Pacing: a technique to manage energy levels and avoid exacerbating symptoms.
- Red light therapy: may help stimulate mitochondria and improve energy production.
- Humming: may help increase nitric oxide levels and improve vascularization.
- Autonomic rehabilitation: gentle, non-exercise-based interventions to help regulate the autonomic nervous system.
- Breath work: gentle, non-exercise-based interventions to help regulate breathing and calm the body.
Existing and promising theories as to why long-COVID is happening:
- Viral persistence: the idea that the COVID-19 virus may persist in the body, causing ongoing symptoms.
- Mitochondrial dysfunction: the idea that the mitochondria, responsible for energy production, may be impaired, leading to fatigue and other symptoms.
- Co-infections: the idea that other viruses, such as Epstein-Barr or CMV, may be reactivated or exacerbated by COVID-19, contributing to long COVID symptoms.
- Microclots: the idea that small blood clots may be forming in the body, leading to inflammation and other symptoms.
- Autoimmune responses: the idea that the immune system may be overreacting or misfiring, leading to chronic inflammation and symptoms.
- Oxidative stress: the idea that the body’s natural antioxidant defenses may be overwhelmed, leading to cell damage and symptoms.
- Post-exertional malaise: the idea that exercise or exertion may trigger a worsening of symptoms, possibly due to the body’s inability to properly recover from physical activity.
These theories are not mutually exclusive, and it’s possible that long COVID is the result of a complex interplay between multiple factors.
🌸 Thu, Aug 1, 2024 🌸
If you are currently implementing or about to implement logging functionality, most likely you are thinking of doing it with a format string, and most likely the compiler would not be able to catch if the format string is not aligned with the parameters, and if the parameters are not aligned, the app will crash. So, please don’t do it. – A developer in pain.
You can do it in the streams C++ style std::cout << 1 << 2 << 3; or you can do it in a comma-separated style LOG(1, 2, 3).
🌸 Thu, Aug 1, 2024 🌸
Yesterday, I had some dried Thai chili in my meal; it tasted delightful, and I had some thoughts: “probably will regret it later.” But I discarded that thought and continued to enjoy the meal. Today, I’m paying the price for yesterday’s enjoyment. 😄
Warnings in C++ are errors. Treat them as such, because they can lead to unexpected behavior or crashes. The code below gives you only a warning, and by the C++ standard, it is undefined behavior for 32-bit integers.
int n = 1e10; // undefined behavior for 32-bit int
I have a call of void situation quite often or intrusive thoughts, then I am watching the progress bar for 2 hours and at 99% I want to press the cancel button really really badly. Never did it, but oh boy!
LLaMA 3.1 is very bad at code completion, not sure exactly why, most likely I broke something, I’ll debug it later.
Yesterday, I spent some time upgrading from LLaMA 3.0 to LLaMA 3.1. I also updated the inference engine, llama.cpp, which was 3 months behind.
So far, the 8B parameters model is way faster and smarter. The 70B model barely fits in my GPU memory and is very slow. I don’t think it’s useful for real-time tasks, but I can batch some tasks and run them overnight.
In the morning, I copy-pasted some questions I had in the past to ChatGPT, and surprisingly, some answers were even better with the 8B params model - sic! That’s amazing.
TLDR; it’s faster and smarter, but the 70B parameter model is a little too big.
So far, Llama 3.1 is pretty good. It has some hiccups. It feels like it’s a bit worse on coding tasks, but it’s hard to judge after a couple of hours of using it.
I found stylized assets pretty much useless, if you are making a stylized game, most likely stylized assets' style would not match with your game’s style. I think if you aim to sell your assets, it is better to make realistic assets.
My laptop screen fell off because it was apparently just glued with double-sided tape; I just put it back and everything works fine. If that happened with my Apple laptop, it would go to the trash or cost me a fortune to fix. Honestly, I prefer the first scenario, where I just slap it back and it works perfectly fine.
I am wondering why client developers' style of programming is so different from server (backend) style. I am looking at the client code, and it is C++; however, it feels alien and wrong. Global variables, nested if statements, awkward interfaces, singletons everywhere, and multiple layers of proxy classes.
Udio just released v1.5, one huge thing they improved is audio quality. I have mixed feelings of excitement and feeling behind all the technological advancements.
I was wondering how many times I should say something or message something until people would, with a 95% chance, get it and not be annoying? The information was in the JIRA ticket, and I explicitly mentioned it, and still people missed it. What is the sweet spot - five times?
Normally, I just keep the tab open and at some moment I get back to it, but with YouTube videos, it is annoying that it just starts autoplaying as soon as I open the tab, so I add the video to the watch later playlist and get back to it in 2 months, then I completely forgot why I did it. It would be nice to be able to disable autoplay or add the ability to add a note to the video, to explain why I added the video to the watch later.
I think it’s okay quality for voice acting to use Voice Changer Client and RVC models. Anyway, currently I don’t have a budget to hire voice actors.
I just cannot keep up with everything that is happening in the world: Donald Trump assassination attempt, Biden drops from the race, CrowdStrike issue, secret service director steps down, and more and more and more…
Shell I write a DAW from scratch instead of making games? LOL
Did you know you can use ChatGPT 4o without a subscription? I just canceled mine.
My whole life, I’ve been told that if you believe in something, it will become true. However, my life experience tells me that no matter how much you believe, if it’s false, it will remain false. If someone gives you a shiny yellow piece of metal and tells you it’s gold, but it’s actually copper, no matter how strongly you believe it’s gold, it will never become gold. I think I should trust my life experience, not what people tell me.
I am considering starting to brand my games under “One Electron Universe studio”. I bought two domain names: 1eu-games.com and 1eu.games. I think they are awesome domain names.
It took 35 minutes just to actually start working. Windows updates, some connectivity issues, reboots. Hey, no complaints, I have been paid for that downtime.
It is always stumbling for me to know how I should write a float zero literal: 0.f or .0f. Probably I’ll do it as 0.0f to be symmetrically unhappy.
It is good that we have a bit of competition in the LLM space: ChatGPT, Meta.ai, Claude.
Yesterday I watched some interviews about LongCOVID, one thing I learned is that the more effort you put into getting out of the condition, the deeper you are getting into it, it’s like quicksand or non-Newtonian fluid. The way to get out is to pace your way out.
Previous - 1 - 2 - 3 - 4 - 5 - 6 - 7 - 8 - 9 - 10 - 11 - 12 - 13 - 14 - 15 - 16 - 17 - 18 - 19 - 20 - 21 - 22 - 23 - 24 - 25 - 26 - 27 - 28 - 29 - 30 - 31 - 32 - 33 - 34 - 35 - 36 - 37 - 38 - 39 - 40 - 41 - 42 - 43 - 44 - 45 - 46 - 47 - 48 - 49 - 50 - 51 - 52 - 53 - 54 - 55 - Next