Skip Navigation

Posts
21
Comments
759
Joined
2 yr. ago

Reddit sucks

  • Haha perfect expressions

  • Thanks! So any gguf file should be safe? I’ve been downloading them from huggingface.

    Yeah it’s wild what some people are letting models do with MCP. Really the Wild West.

  • Thanks! Is Qwen considered trustworthy?

    I’ll check out a network sniffer.

  • For sure. You would need a model that is not censored at training.

  • At a high level I think it’s similar to integration of any other vendor. ROI calculation based on cost of vendor and value of costs reduced and/or revenue gained post integration. Companies may have some granular attribution models to say X investment in AI project was directly tied to Y outcomes valued at $Z.

    Some execs are admitting that AI is costing more than humans.

  • The reasoning for hello is crazy haha. I’ve experienced the same, but if you turn off reasoning on launch and explicitly state the rules you want it to break I’ve had some success. I was trying to get it to tell me a story about llamas having sex and it went on forevvver reasoning about why it shouldn’t say things and how to rephrase to not break rules. The funniest part of the reasoning was “llamas don’t have penises (obviously, they’re mammals)”. Haha it reasoned itself into thinking llamas, and mammals, don’t have penises.

  • How does the model connect to the internet if I don’t give it a tool to? What if I’m not connected to the internet while using? Does it then send the packets after I connect? Is this documented somewhere? What’s a better model that doesn’t do this?

  • Thanks! I don’t think I can run an 8B yet. Need to invest in a better machine. I’m stuck on 4B Q4.

    The uncensored Qwen that I’m using started throwing infinite ?’s at me one time. Had to restart it and has been fine since.

  • Jesus that MoE wiki is a fucking rabbit hole.

    Thanks for sharing! Unfortunately I haven’t invested in a decent computer yet. Using 16GB GPU so been stuck on 4B Q4’s.

    I’m not particularly interested in ERP, but I have obviously been using it for testing models. I’m more curious about other topics with guardrails.

    I noticed that Qwen 3.5 uncensored is good if I turn off reasoning and explicitly say I want it to break the rules.

    I’ll check out sillytavern tho. Thanks!

  • Thanks for the explanation!

    The use case is writing marketing communications to match a library of content that a company has already written.

    We’re currently using RAG and it’s okay, but I’m wondering how much better it would be if it were tuned.

  • Thanks! I’ll check out that model. Is it actually usable or just good at being uncensored?

  • Thanks! I’ll try it out. I’m on an old phone and resistant to switch to a bigger one.

  • Removed Deleted

    Permanently Deleted

    Jump
  • That’s funny. I actually think screens on some appliances are useful, like coffee/espresso machines. There are so many setting on some that it’s much easier with a screen. I guess it also depends if you use additional features or not. E.g. schedule something to do something with specific settings

  • Thanks! I’ll do some research.

  • Thanks! That’s interesting that RAG alone would be better than a tuned model. Why is that? What If you have a very specific task, like writing copy based off existing documents and decisions are based on a set of specific variables?

    What if you use RAG and tune it? Any benefit there?

    Last question. Would a fine tuned model be more energy efficient than a model using RAG?

  • Thanks! I’ll do some reading.

  • Thanks! I’ll do some experimentation.

  • Ah yeah I’m not trying to do all that. Is there a way to keep your computer almost asleep just waiting for a signal?

  • Thanks! Ik interested in Wake on LAN. I don’t wanna keep drawing power when I’m not using it. Any recommendations for setting that up?