Open Source AI is Amazing

D

dogememe

Joined
Apr 15, 2017
Messages
6,591
Location
California
A year ago, self-hosted AI needed so much hand holding and made tons of mistakes that it was practically useless.

Now, I vibe-coded an entire PHP-based content management system in a few hours using Qwen3.6 running on my AMD R9700s. On a self-hosted AI server management/dashboard thing that I also mostly vibe-coded.

I just hope we get Qwen3.7 soon. They apparently got rid of most of their open source team though so IDK.
 
I have a similar setup. It is amazing what you can do with models in the 1b to 250b (MOE) parameter range. They’re not as good as the giant commercial ones but if you know what you’re doing they can be very useful.

I use them for TTS, image gen, and even some agentic automation tasks. Coding… I prefer to keep on the big models, personally. I just find they have a better grasp of the big picture, even if I assign a large context window.
 
DCDG said:
BITOG doesn't care much for AI
Click to expand...
I like it, and I'm one of the longest time members here.

Note: Prior to retirement I did work for the father of the internet, founder of 3D graphics and yes AI. So while not on the 'inside' so to speak as I simply worked in the flight department. Our flight department did fly Peter Thiel around during the early days of Palantir. I still visualize him as a very young guy! Super nice guy too!!!
 
goingplacesanddoingstuff said:
I have a similar setup. It is amazing what you can do with models in the 1b to 250b (MOE) parameter range. They’re not as good as the giant commercial ones but if you know what you’re doing they can be very useful.

I use them for TTS, image gen, and even some agentic automation tasks. Coding… I prefer to keep on the big models, personally. I just find they have a better grasp of the big picture, even if I assign a large context window.
Click to expand...
Agreed, the frontier models are better at understanding the entire project rather than just implementing a particular thing (for example, if you're not careful and ask Qwen3.6 to do some front end design work on a particular feature, it'll often make something that looks nice but with no regard for the styling of the entire site/app).

The way I get around this is being more careful with how I prompt it and what I ask for. Plan made in whatever agent/IDE you use is very handy. Tell it what you want, then review the plan to see if it got something wrong or made an assumption. Then work off that. Without that you'll waste 5 minutes waiting for it to do something and then it didn't do what you want.

Although I run multiple models concurrently and am usually doing multiple things at once so I am rarely waiting for it. And if I need something really fast the $16/mo Cursor plan is good. I just open that up, use Composer 2.5 fast, which zooms.
 
Cujet said:
I like it, and I'm one of the longest time members here.

Note: Prior to retirement I did work for the father of the internet, founder of 3D graphics and yes AI. So while not on the 'inside' so to speak as I simply worked in the flight department. Our flight department did fly Peter Thiel around during the early days of Palantir. I still visualize him as a very young guy! Super nice guy too!!!
Click to expand...
The haters are Pavlovian at this point. Completely unidimensional, pre-programmed reactions based on a limited understanding of the technology.

If they don’t get it by now they’re probably never going to.
 
DCDG said:
BITOG doesn't care much for AI
Click to expand...
It's more about having rules of how AI is cited or used around here. BITOG has no qualms with people using AI for their own usage.

AI is here to stay and isn't going away. It's best to avoid being a luddite and instead focus on how to function best in a world with it. For some of us, AI is nearly a superpower of productivity enhancement, a dream come true.
 
Very limited. The report by Goldman Sachs that came out a few years ago that got headlined & thrown around the interwebs a while back, about how 300 million jobs would be replaced by AI, is a classic example. The report actually stated that every major increase in technology (such as the use of the steam engine, development of cars & computers) did supplant some jobs, but created many more. But that part gets left behind with all the hand-wringing about AI going on.

goingplacesanddoingstuff said:
The haters are Pavlovian at this point. Completely unidimensional, pre-programmed reactions based on a limited understanding of the technology.

If they don’t get it by now they’re probably never going to.
Click to expand...
 
paulri said:
Very limited. The report by Goldman-Sachs that came out a few years ago that got headlined & thrown around the interwebs a while back, about how 300 million jobs would be replaced by AI, is a classic example. The report actually stated that every major increase in technology (such as the use of the steam engine, development of cars & computers) did supplant some jobs, but created many more. But that part gets left behind with all the hand-wringing about AI going on.
Click to expand...
Yep. They act like if we solve every problem we have right now then there are no more problems left. Yeah, it doesn’t work like that.
 
AI will never pull wire through a 125 degree attic or re plumb an entire house. but yeah I get it it's cool to help a programmer that's not really good at programming or maybe run the robotics for my next colonoscopy.
 
paulri said:
Very limited. The report by Goldman Sachs that came out a few years ago that got headlined & thrown around the interwebs a while back, about how 300 million jobs would be replaced by AI, is a classic example. The report actually stated that every major increase in technology (such as the use of the steam engine, development of cars & computers) did supplant some jobs, but created many more. But that part gets left behind with all the hand-wringing about AI going on.
Click to expand...
AI will reduce some jobs in the legal field. If properly implemented, my wife's job would be a lot less in demand. She was a legal secretary and AI can write legal complaints, motions, etc a lot faster than she can type them out. Currently, her firm has one legal secretary per attorney. That could be reduced to one legal secretary per four or five attorneys. Her firm is very Luddite, but even they are looking at AI implementation with, from what my wife understands, a local AI.
 
I'm not going to say that AI can go through your attic. That is actually proving a point that many are saying about AI, that it isn't going to replace everybody's job.

But it surely has many capabilities. Take this article from WaPo, talking about how the US military used AI in the strikes on Iran (not sure if this is behind a paywall or not, I believe this article can be gifted for the next 14 days):

In the American military, the system allowed one artillery unit to do the work of 2,000 staff with a team of just 20 people, according to a study of the system’s use by the Army’s 18th Airborne Corps by Georgetown University.
Click to expand...
The "system" referenced in the quote above was the US military's Maven Smart System, which was according to the article, built by Palantir & powered by Claude, whatever that is supposed to mean.

DCDG said:
AI will never pull wire through a 125 degree attic or re plumb an entire house. but yeah I get it it's cool to help a programmer that's not really good at programming or maybe run the robotics for my next colonoscopy.
Click to expand...
 
Last edited:
I'm sorry dogememe, I can't do that.

1782396384172.webp
 
I don’t hate AI. I pay a monthly fee OOA $200 to ChatGPT, but use AI little.

What I would see as a concern of AI is a continuous dumbing down of people’s god given gift of researching, and critical thought and thinking.

AI surely makes me lazy, and gives information I likely take as fact, and the information may be absolutely false.


Kind of like video games of sports replacing actual sports. One misses out on the leadership, followership, pressure, and preparation one receives when playing actual sports, and receives none of those benefits playing video game sports.
 
GON said:
I don’t hate AI. I pay a monthly fee OOA $200 to ChatGPT, but use AI little.

What I would see as a concern of AI is a continuous dumbing down of people’s god given gift of researching, and critical thought and thinking.

AI surely makes me lazy, and gives information I likely take as fact, and the information may be absolutely false.


Kind of like video games of sports replacing actual sports. One misses out on the leadership, followership, pressure, and preparation one receives when playing actual sports, and receives none of those benefits playing video game sports.
Click to expand...

No judgement but if you don't use AI much why pay $200/mo for ChatGPT? In two years that's $4800 and you could have built a sweet self-hosted AI server.
 
paulri said:
I'm not going to say that AI can go through your attic. That is actually proving a point that many are saying about AI, that it isn't going to replace everybody's job.

But it surely has many capabilities. Take this article from WaPo, talking about how the US military used AI in the strikes on Iran (not sure if this is behind a paywall or not, I believe this article can be gifted for the next 14 days):


The "system" referenced in the quote above was the US military's Maven Smart System, which was according to the article, built by Palantir & powered by Claude, whatever that is supposed to mean.
Click to expand...
Operation Iraqi Freedom they were putting scud missiles in bed with the bad actors before AI. lol
 
You must log in or register to reply here.

Similar threads

D
AI at home, round 2
Replies
18
Views
346
dogememe
M
1 Week Review - 2025 Toyota Tacoma Hybrid Limited
3 4 5
Replies
96
Views
35K
meep
S
  • Locked
Kicking the Microsoft Habit
Replies
15
Views
4K
ToyotaNSaturn
G
  • Locked
Anything into oil (from Discover Magazine)
2 3
Replies
51
Views
37K
'Tard
Back
Top Bottom