Title: breaker breaker still doing maintenance and informational research

2025-01-13 20:37:39 +0800

-----

2025-01-07 18:03:58 -0800

Rod Dreher talking about occultism in ai industry

2025-01-08 04:23:14 -0800

still downloading this thing. 80% now. Got a function to attend this afternoon. notes here will be sparse, if there are any at all.

2025-01-09 09:31:11 -0800

llama 3.3 is downloaded. now to see if I can figure out how to run it.

2025-01-09 10:13:16 -0800

figured it out. fans kick on as soon as I ask it a question. very slow, too. I guess the web-based llms are running on well endowed cloud platforms. in any case, I downloaded llama3.3 because it's got 70 billion paramaters and I thought the bigger the better. but it took me only a few minutes to find out that this thing is severely guardrailed and wont produce more than ten lines of code.

so after a little poking around I found out about lm-studo and there's a model on huggingface called Wizard-LM-7B-Uncensored that looks like it's just a plain denerfed llm though I dont know what the foundational model is, or if it's a single foundational model or even how any of that works. I got a lot of shit to figure out. which is good because the material is right at hand and it shouldnt take me more than a few days to onboard some amount of new marketable capability.

2025-01-09 11:53:36 -0800

where did the time go. I managed to get LMStudio running after fussing with it for a while. apparently a long while. now downloading wizardlm-7b-uncensored-gguf because I want an llm that will say 'fuck'. what's the point of running an llm locally if it's going to be all pansified like the web-based ones. Ill also need a model that can code well. how I'm supposed to use it with cursor, I'm not sure.

anyway, an hour till this model is downloaded so I'm going to do some more cleaning.

2025-01-13 04:31:17 -0800

almost five full days since I've sent one of these posts out. nearly four full days since I've even made an entry. done as much downloading as seems necessary right now, but I'm guessing it's 50/50 whether I need to download more once I start messing around with the tech again. while the downloading has been going on, and I've been keeping off the computer to keep the tiny little download pipe as clear as possible, I've gotten deeper into getting the building cleaned up as well as deepering into the background research. I'm actually running through some online courses that have some info I want. there's a bit of that left. it could probably take up a lot more of my time if I wanted it to. in any case, there are a few things I want to wind up before I get back to the direct ai research so entries here might continue to be sparse. I will send this out now. it seems full enough and it's been too long since I've sent anything.