Posts by W8n4Singularity

1) Message boards : Cafe : ML & more in the news (Message 1324)
Posted 17 Aug 2021 by W8n4Singularity
Post:
Interesting way to save computation cycles or pondering?
https://arxiv.org/pdf/2107.05407.pdf
2) Message boards : Science : Artificial intelligence speeds forecasts to control fusion experiments (Message 1291)
Posted 28 Jul 2021 by W8n4Singularity
Post:
Any general insights that are gained from this project could be utilized to build better AI or even just to analyze it better. We haven't even heard how different researchers will analyze and compare MLDS. I believe any knowledge gained from this project could benefit all AI research.
3) Message boards : Cafe : GitHub and OpenAI launch a new AI tool that generates its own code (Message 1242)
Posted 1 Jul 2021 by W8n4Singularity
Post:
From the Verge: GitHub and OpenAI launch a new AI tool that generates its own code
Official website: OpenAI Copilot

This tool looks amazing for trying to write complex code. I have seen projects on Youtube using AI to write code from a comment, but this looks like it will soon be fully released to the public.

I also can't help but think that this type of technology will be a huge help in reaching AGI. Not just as a direct result of using the tool to help write complex code, but also by helping amateur coders to learn to code easier.
4) Questions and Answers : Issue Discussion : Out of work for CPU (Message 1235)
Posted 16 Jun 2021 by W8n4Singularity
Post:
From what I understand DS3 is what is currently running on CPU; DS1 and 2 are currently GPU tasks. The latest news article mentions DS4 will be ready any day now. I am simulating some black holes in the meantime.
5) Message boards : Cafe : New to the project (Message 1222)
Posted 11 Jun 2021 by W8n4Singularity
Post:
Welcome! I completely agree. In the 6 years I have been running BOINC, this project is by far my favorite for that reason, plus the potential for improving our understanding of AI. Everyday I see new research showing more and more complex AI that we understand less and less. Without projects like this, AI is doomed to be narrow and poorly understood.
6) Message boards : News : [TMIM Notes] June 8 2021 (Message 1219)
Posted 11 Jun 2021 by W8n4Singularity
Post:
Fantastic job! I am staying 'til teal! I am already looking forward to learning more about DS4 and the next project (personally I hope it targets reproducibility).
7) Questions and Answers : Unix/Linux : Linux for Windows Users (Message 1210)
Posted 31 May 2021 by W8n4Singularity
Post:
Try updating the GPU drivers in the OS settings in the GUI. Otherwise I am not sure, you can always give Ubuntu a try. Wish I could be more help.
8) Questions and Answers : Unix/Linux : Linux for Windows Users (Message 1202)
Posted 24 May 2021 by W8n4Singularity
Post:
I prefer Linux Mint, although Ubuntu is always a good choice. If you are used to Windows, Linux Mint is the closest and is based on Ubuntu.
9) Message boards : News : New client released with ARM support! (Message 1142)
Posted 8 Apr 2021 by W8n4Singularity
Post:
I know this is an old post, but I have been thinking about Android support lately and am curious if an Android client is currently feasible. I have ran BOINC projects on multiple Android devices over the years and personally highly support the idea. Unfortunately, there are large barriers that would prevent many users from participating, aside from development issues and statically linking pytorch. This is the fact that the BOINC app for Android is so outdated that it does not allow users to add newer projects or account managers using URL.

I am fully aware there are work-arounds, such as using one of the 3 already included account managers, using an edited version of the APK, or using a third party BOINC client. Unfortunately, even the edited version of the APK is so old it will not run on certain devices and using an account manager is not optimal for many peoples set-up. I also reached out to the BOINC android developer last year who said they are aware of the issues and were awaiting help from someone with Android development. That was about a year ago, so I am not sure if there are any updates on that project.

This is just my 2 cents, thank you for reading!
10) Message boards : Cafe : How does MLC verify results without running multiple tasks per work unit (redundancy)? (Message 1141)
Posted 8 Apr 2021 by W8n4Singularity
Post:
Very interesting. If that is the case, does that also mean that MLC would need to process far more results to reach statistical significance versus a project that uses redundancy and can prove that the calculated results are valid?
11) Message boards : Cafe : How does MLC verify results without running multiple tasks per work unit (redundancy)? (Message 1136)
Posted 7 Apr 2021 by W8n4Singularity
Post:
Asking just out of curiosity.

Some older tasks did have other canonical results such as this one: https://www.mlcathome.org/mlcathome/workunit.php?wuid=1120016
Newer tasks are validated without redundancy: https://www.mlcathome.org/mlcathome/workunit.php?wuid=2602952

Thanks!
12) Message boards : News : [TWIM Notes] Feb 23 2021 (Message 1119)
Posted 8 Mar 2021 by W8n4Singularity
Post:
I just noticed a substantial increase in the run time of the rand-automata units on my CPU, going from <7 hours to >9.5 hours. Was the length of these tasks increased again? Not an issue at all, just curious. Thanks!
13) Message boards : News : [TWIM Notes] Feb 23 2021 (Message 1108)
Posted 26 Feb 2021 by W8n4Singularity
Post:
Keep up the amazing work! Personally, I really like the concept of Gridcoin and can see whitelisting bringing in more contributors.
14) Message boards : Cafe : Geeking out on emerging technologies ... (Message 870)
Posted 20 Nov 2020 by W8n4Singularity
Post:
I believe AI, such as IBM's Watson, will soon drastically improve diagnostic accuracy and treatment optimizations in healthcare. While AI is nowhere near replacing a human doctor, giving a doctor a diagnosis, treatment, and reasoning for such, can make a big difference. In addition to giving a recommendation, AI is also very important in healthcare due to its ability to come to a conclusion based on obscure data in medical records. A patient in the hospital for just a few days can generate hundreds to thousands of pages of data (lab values, vitals, visit reports, SOAP notes, etc). Personally, I can't wait for better AI tools and do not believe that AI will threaten my job.
15) Message boards : News : Badges! (Message 865)
Posted 20 Nov 2020 by W8n4Singularity
Post:
Maybe the 500M and 1B goals will persuade a few people to buy a supercomputer!
16) Message boards : News : [TWIM Notes] Oct 19 2020 (Message 690)
Posted 25 Oct 2020 by W8n4Singularity
Post:
Great work! As for the GPU tasks, I am excited to try. However, I do think it would be nice to have a "no GPU tasks" option in settings for people who want a choice.
17) Questions and Answers : Issue Discussion : Rpi4 now erroring new tasks. (Message 680)
Posted 25 Oct 2020 by W8n4Singularity
Post:
In terms of BOINC, I've had better luck using Ubuntu server 64-bit. Ubuntu server is headless by default, so it uses a lot less RAM compared to having a desktop. I use the open-server SSH function to remotely control the headless BOINC-client.
18) Questions and Answers : Issue Discussion : Rpi4 now erroring new tasks. (Message 674)
Posted 23 Oct 2020 by W8n4Singularity
Post:
I feel your pain, man. I've had my fair share of "fun" with Pi's.
19) Questions and Answers : Windows : Rand_automata time remaining (Message 662)
Posted 20 Oct 2020 by W8n4Singularity
Post:
Not sure about the specifics, but I have seen this behavior multiple times in the past crunching for different projects. I believe this is expected to occur when new applications are added and will fix itself over time.
20) Questions and Answers : Unix/Linux : Pi4 (Message 630)
Posted 9 Oct 2020 by W8n4Singularity
Post:
It seems to works well on RPi4 4GB. Most tasks have a run time of about 7 hours, while the rand_automata tasks take about 31 hours. Very roughly, for the rand_automata tasks, the Pi4 uses about 0.25 kWh for 4 tasks (~0.063 kWh/task) while Ryzen 3950x uses 1.6 kWh for 32 tasks (~0.05 kWh per task). (However irl I run the Ryzen at 75% cores for multitasking).


Next 20

©2021 MLC@Home Team
A project of the Cognition, Robotics, and Learning (CORAL) Lab at the University of Maryland, Baltimore County (UMBC)