segmondy

25 ms·

631.
by segmondy 2y ago
For those that don't know, Firefox already has a built in AI assistant that you can use with your own models or point to external models https://support.mozilla.org/en-US/kb/ai-chatbot
632.
by segmondy 2y ago
I saw about the book earlier, didn't care about reading it. Then I saw about the arbitration on the news and immediately ordered a copy.
633.
by segmondy 2y ago
No, money ruins open and free source. If you are going to do it, do it for the passion. Every single open/free software that I have rooted for that decided that money is important turned out terrible. Please don't ask for a li
634.
by segmondy 2y ago
The question will be how it will perform. I suspect Deepseek, Llama405B demonstrated the need for larger memory. Right now folks could build an epyc system with that much ram or more to run Deepseek at about 6 tokens/sec for a fract
635.
by segmondy 2y ago
Idea is 8 years old https://techcrunch.com/2017/03/07/land-rovers-new-discovery-...
636.
by segmondy 2y ago
It's amazing how good and stable it got really fast. I was able to install it and use it with X and for everything in 1994 while still in high school. It's been my only OS ever since and set the path for my career and life.
637.
by segmondy 2y ago
My main desktop at home is a 2011 hp envy. I sometimes think of upgrading, but why? It works. my main laptop is a roughly 10yrs old thinkpad. Outside of macbooks, PCs haven't really given any reason to upgrade unless one is a har
638.
by segmondy 2y ago
You gotta have super thick skin to be a maintainer of an opensource project or even be popular on the net these days. Folks are going to come for you for whatever reason, if you read too much into it you're going to have a bad time.
639.
by segmondy 2y ago
how much context size?
640.
by segmondy 2y ago
Let me guess, it will drop after the shoe drops from people getting covid vaccines?
641.
by segmondy 2y ago
"Through this incident I've learned people should instead contact their home insurance company and get an approved list of vendors to choose from." I don't agree with this. Never contact your insurance company unless yo
642.
by segmondy 2y ago
That's not what they are running. the dynamic quants are Q1, Q2. The $6000 is running Q8.
643.
by segmondy 2y ago
The answer is yes.
644.
by segmondy 2y ago
Do not think of a pink elephant. Were you able to do so?
645.
by segmondy 2y ago
They pointed out that the cost calculation is based on if those GPUs were rented at $2/hr. They are not factoring in the prior cost of buying those H800s because they didn't buy it to build R1. They are not factoring in the cost
646.
by segmondy 2y ago
Nah, this is not about Nvidia being a bubble. This is about people forgetting that software will keep eating the world and Nvidia is a hardware company no matter how many times people say it's a software company and talk about Cuda.
647.
by segmondy 2y ago
No, this is not the first local model with a context length of greater than 128k, there have been such models, for example the following https://huggingface.co/ai21labs/AI21-Jamba-1.5-Mini 256k https://huggi
648.
by segmondy 2y ago
You can't on modest hardware, VRAM size is a function of model size, KV cache that depends on context length and the quant size of the model and K/V. 16gb isn't much really. You need more vram, the best way for most folks is
649.
by segmondy 2y ago
No one is forcing you to subscribe. You can code the old fashion way, if you wish to use AI, you can run your own local model.
650.
by segmondy 2y ago
If it's truly talking to OpenAI or Claude, then the traffic can be intercepted and redirected to your own AI router for other cloud AI providers or your local Ollama instance.
651.
by segmondy 2y ago
Trae doesn't look opensource, from their site looks like you download the binary (osx only for now) and install it. You can't configure it to use your own AI or another provider. You need to use their own API, all your prompts an
652.
by segmondy 2y ago
The same way you validate it if you didn't use an LLM.
653.
by segmondy 2y ago
You are not pushing it at 100. I can classify "Is 20 bottles of ferric chloride' a service or product in probably 2 seconds with a 4090. Something that most people don't realize is you can run multiple inference. So wit
654.
by segmondy 2y ago
Yup, I did an experiment a long time ago, where I wanted best of 2. I had Wizard, Mistral & Llama. They would generate responses and I would pass the response to all 3 models to vote. I would pass it in to a new prompt without refer
655.
by segmondy 2y ago
They lose their customers. TSMC succeeded because they didn't compete with their customers. When they started, they focused on making chips for others. Any other company that that had a Fab could offer the same service at the same q
656.
by segmondy 2y ago
wyze sells their new robot vacuum for about $100-$150 every year, you have to look around nov - dec (black friday and christmas sells), you might also find it sold by walmart for $100. it usually goes for around $300.
657.
by segmondy 2y ago
tell us how to do it right.
658.
by segmondy 2y ago
yup, seems sensible because we have been getting gouged to death with high prices for a minute. Being that 4090's are going for $2000. I suspect we are going to see 5090's easily settle at $3000. The demand from the AI space
659.
by segmondy 2y ago
ebay is still useful for obscure stuff. if you want to buy used servers, GPU, electronics, car parts, etc. I don't think ebay and FB MP really overlapped, MP ate craigslist. The closest thing to eating ebay is Aliexpress.
660.
by segmondy 2y ago
I did experiment with local LLM running on my machine, most solutions were generated within the time of 30-60seconds. My overhead was really copying part 1 of the problem, generating the code, copying and pasting the code, running it, cop

More ›