e-beach
10 MW
Dang!!! I was waiting for an email or something seeking a paypal interaction or something.......
Endless Sphere knowledgebase system ( zeropress ) and funding drive
- Thread starter neptronix
- Start date
If you still want to donate i can open it back up, you let me know!
I did ask for an amount where we're cutting corners a bit to make it happen. If you still want to donate so we can do a bit better, i can open up the ko-fi fundraiser again.
I did ask for an amount where we're cutting corners a bit to make it happen. If you still want to donate so we can do a bit better, i can open up the ko-fi fundraiser again.
e-beach
10 MW
If you open it up today I will donate $100 this afternoon.
Fundraising opened again, thank you!
https://ko-fi.com/es_kb_fundraiser
https://ko-fi.com/es_kb_fundraiser
e-beach
10 MW
PaulD
1 kW
I also missed the fundraising. Is more funding helpful at this point? If so, how much?
It would be helpful, i'm in the process of finishing up the design ( visual and logic ) and thinking the budget is a little tight.
You pledged $500, but i think $250 would be fine and give us some headroom in the labor hours. If i'm wrong, the remainder becomes a moderator tip jar.
You pledged $500, but i think $250 would be fine and give us some headroom in the labor hours. If i'm wrong, the remainder becomes a moderator tip jar.
Dui ni shuo de dui
100 kW
Dude, the same thing happened to me last week (account locked).
It was just because my password was a bit too short. The only thing you need to do is to change it for a longer one, it's not rocket science.
As you can see, I got my account back, it wasn't a big deal. And I wasn't even a fancy moderator.
So I suggest you stop being salty for no reason, you take two minutes to update your password with a new one and everything will be fine again. Easy.
Also I don't see anything wrong with asking for people's financial help to develop new features for the forum, it's not like anyone is forcing you to give anything.
i will do 50$
We're over our target for the non-luxury version and happy with the fluff money we got. to account for the fact that software developers are awful at estimating money 
Everyone done being generous? i should shut the funding pipe sooner than later.
Everyone done being generous? i should shut the funding pipe sooner than later.
Build progress thread here:
Zeropress / ES Knowledge base progress thread
Zeropress / ES Knowledge base progress thread
Hey yall.
I want to say thank you again for your previous donations. We are beaming that we got to make this addon to the forum possible.. and also.. that we're ~95% done at this point!
Some WIP screenshots:
Designing a visual edit history turned out to take the project way over budget. We tried a HTML code-only edit history, but it sucked to use in practice. The visual edit history added an additional month of work we didn't expect. And there's a little more work to do to refine things remaining.
Retaining the previous 3 way split, to finish the project, we could use another $750 from the forum to finish the software. Sorry to ask again, but can you pitch in?
Donations can be made here: https://ko-fi.com/es_kb_fundraiser
I want to say thank you again for your previous donations. We are beaming that we got to make this addon to the forum possible.. and also.. that we're ~95% done at this point!
Some WIP screenshots:
Designing a visual edit history turned out to take the project way over budget. We tried a HTML code-only edit history, but it sucked to use in practice. The visual edit history added an additional month of work we didn't expect. And there's a little more work to do to refine things remaining.
Retaining the previous 3 way split, to finish the project, we could use another $750 from the forum to finish the software. Sorry to ask again, but can you pitch in?
Donations can be made here: https://ko-fi.com/es_kb_fundraiser
jonescg
100 MW
marty
10 MW
marka-ee
100 W
- Joined
- Mar 24, 2020
- Messages
- 252
It feels like old-fashioned technology to me. Have you looked into large language models and RAG? It's the way companies typically train a large language model so that it's expert on their particular database of documents and information.
Also, this technology doesn't require much money to implement. It's all open source.
What the result would be is something similar to chat GPT where people can ask questions and it would reply in a fairly informative way because it was trained on all the data from the source. Of course, this wouldn't provide pictures. It would be text only.
Also, this technology doesn't require much money to implement. It's all open source.
What the result would be is something similar to chat GPT where people can ask questions and it would reply in a fairly informative way because it was trained on all the data from the source. Of course, this wouldn't provide pictures. It would be text only.
Yeah but we're not a big tech company
I've actually thought about this because google search quality keeps dropping as the years go by.
Providing a LLM search on our existing 8gb of text in the database would cost ~100 times our current $80/mo cost we pay.
Example: to run a typical query on my Nvidia 4070 with qwen2.5-14b, i'm typically using 15 seconds of 125W extra power. Multiply that by 50 to serve the forum. The numbers don't look pretty.
This 8gb of text would have to be hand curated and we certainly can't afford to pay for that labor or expect people to perform the labor for free. Right now, Open AI pays a fleet of Kenyans $1/hr to do that.
Once AI costs get out of the stratospheric range, it's worth considering.
The knowledgebase we've built, if utilized to it's potential, is maybe tends-hundreds of megabytes of text in total, since it's focused on knowledge, and not just discussion + knowledge. It might be fiscally feasible some day to provide search from this much smaller, and already cherry picked dataset.
The spiritual purpose of zeropress is to make the production and consumption of knowledge as fast and efficient as possible. Adding AI search does serve that purpose.
Anyway, the thing is >99% built, no turning back now
I've actually thought about this because google search quality keeps dropping as the years go by.
Providing a LLM search on our existing 8gb of text in the database would cost ~100 times our current $80/mo cost we pay.
Example: to run a typical query on my Nvidia 4070 with qwen2.5-14b, i'm typically using 15 seconds of 125W extra power. Multiply that by 50 to serve the forum. The numbers don't look pretty.
This 8gb of text would have to be hand curated and we certainly can't afford to pay for that labor or expect people to perform the labor for free. Right now, Open AI pays a fleet of Kenyans $1/hr to do that.
Once AI costs get out of the stratospheric range, it's worth considering.
The knowledgebase we've built, if utilized to it's potential, is maybe tends-hundreds of megabytes of text in total, since it's focused on knowledge, and not just discussion + knowledge. It might be fiscally feasible some day to provide search from this much smaller, and already cherry picked dataset.
The spiritual purpose of zeropress is to make the production and consumption of knowledge as fast and efficient as possible. Adding AI search does serve that purpose.
Anyway, the thing is >99% built, no turning back now
Last edited:
A-DamW
1 kW
What! sunken costs and all? I'm in for another $40
Yep!
No major remaining issues, the only thing is that you can't enter PHP/JS/HTML into a code element without it getting bodged ( fault of our WYSIWYG editor, which we have a support ticket in for ), but this isn't a huge problem for ES. So we are ready to update the preview and let people kick the tires.
Google just keeps getting worse. Their Ai generated answers are really bad. I searched "ebike motor winding count" and this is what Ai generated:Yeah but we're not a big tech company
I've actually thought about this because google search quality keeps dropping as the years go by.
Providing a LLM search on our existing 8gb of text in the database would cost ~100 times our current $80/mo cost we pay.
Example: to run a typical query on my Nvidia 4070 with qwen2.5-14b, i'm typically using 15 seconds of 125W extra power. Multiply that by 50 to serve the forum. The numbers don't look pretty.
This 8gb of text would have to be hand curated and we certainly can't afford to pay for that labor or expect people to perform the labor for free. Right now, Open AI pays a fleet of Kenyans $1/hr to do that.
Once AI costs get out of the stratospheric range, it's worth considering.
The knowledgebase we've built, if utilized to it's potential, is maybe tends-hundreds of megabytes of text in total, since it's focused on knowledge, and not just discussion + knowledge. It might be fiscally feasible some day to provide search from this much smaller, and already cherry picked dataset.
The spiritual purpose of zeropress is to make the production and consumption of knowledge as fast and efficient as possible. Adding AI search does serve that purpose.
Anyway, the thing is >99% built, no turning back now
"A typical ebike motor winding count can range anywhere from around 10 to 30 turns per coil depending on the motor's desired characteristics, with higher turn counts providing more torque at lower speeds and lower turn counts offering higher top speeds, but requiring more current to achieve the same torque; however, this is just a general range and exact counts will vary greatly depending on the specific motor design and manufacturer. "
They footnoted 3 articles, one being from Grin, making it seemingly credible, but that article was comparing 6T and 11T motors. The other two articles weren't even ebike related.
I always use at least two search engines now, since they all are commercially driven with respect to search results and order. I realized that a couple years ago when I went to get a brake job, and the guy asked me how I heard of him. I said, my usual guy retired, so I just did a Google search. He told me how all that stuff works, and why his shop turned up at the top. Every search engine does it, and not just the stuff marked as an ad or sponsored at the top.
Yeah, accuracy is still a problem i don't like.
OpenAI recently demoed a modified version of o3 that thinks longer. People were impressed that it could solve a simple puzzle that was hard to solve by AI before. What wasn't mentioned is that puzzle solve costed something like 5-50 gallons of gas worth of electricity. Increased accuracy and ability is very expensive at the moment.
I think progress in AI might be stymied by the end of moore's law and right now, companies are burning money on it just so that they can be first to get a dominant market position.
Something to revisit in a few years as the technology advances.
OpenAI recently demoed a modified version of o3 that thinks longer. People were impressed that it could solve a simple puzzle that was hard to solve by AI before. What wasn't mentioned is that puzzle solve costed something like 5-50 gallons of gas worth of electricity. Increased accuracy and ability is very expensive at the moment.
I think progress in AI might be stymied by the end of moore's law and right now, companies are burning money on it just so that they can be first to get a dominant market position.
Something to revisit in a few years as the technology advances.
Last edited:
marka-ee
100 W
- Joined
- Mar 24, 2020
- Messages
- 252
2025 will bring some big advancements in AI, I think. The use of agentics in AI for more advanced reasoning will bring another level of accuracy and intelligence to query responses.
To see some of the level of excitement, I would recommend watching Jensen's NVIDIA CES opening talk on CES 2025, where he covers a lot of their plans and advancements.
Personally, stepping back and looking at what's happening, I think Jensen will be regarded as the enabler and harberer of the start of the destruction of mankind. This is not going in a good direction for the way humans should live, I believe. Although I think the technology is sexy and as a technological person, I study it closely.
To see some of the level of excitement, I would recommend watching Jensen's NVIDIA CES opening talk on CES 2025, where he covers a lot of their plans and advancements.
Personally, stepping back and looking at what's happening, I think Jensen will be regarded as the enabler and harberer of the start of the destruction of mankind. This is not going in a good direction for the way humans should live, I believe. Although I think the technology is sexy and as a technological person, I study it closely.
I think there will be some disbelief in future that we were able to achieve so much with so little.
Thinking about some great tools from the 'desktop era', that became obsolete and were replaced with new, modern, shiny web 2.0 tech. And only the old users remember how good the old tools were, how much easier and logical the UI was to use and how much has been lost by migrating to dumbed down and simplified new tools.
Not mentioning the fact that the old software could run happily on much less capable computers. And with new AI tools you'll be keeping thousands of computers busy and burning thru tons of electricity to just to be able to write a small bit of code or paragraph of text...
Thinking about some great tools from the 'desktop era', that became obsolete and were replaced with new, modern, shiny web 2.0 tech. And only the old users remember how good the old tools were, how much easier and logical the UI was to use and how much has been lost by migrating to dumbed down and simplified new tools.
Not mentioning the fact that the old software could run happily on much less capable computers. And with new AI tools you'll be keeping thousands of computers busy and burning thru tons of electricity to just to be able to write a small bit of code or paragraph of text...
Last edited:
Similar threads
