Huzzah! Milestone reached, TWhbP (as seen by Google) has had 100 unique visitors (I did discuss this here where I mention some people are invisible to Google so it's likely higher). But this is pretty good progress, I added Google Analytics (GA) on Nov 17, find below a look at the users I see in the GA portal for TWhbP. So ~3 weeks and made it to 100, that's just shy of 5 unique people visiting a day. I think the retention numbers are low, but I've also mentioned that a lot of data is created on the net each day, so it's inspiring that word of mouth and SEO is helping. The few times I've commented on Reddit is in threads I should unsubscribe to, but see a lot of similar rhetoric about the future of the working class. I've stated this might not be for you today, it might not be for you tomorrow. Maybe I'm entirely off key, but like it's sage advice to lay down roads today so tomorrow our children can get to school is the claim I'm making.
I did discuss my blog with a few friends last night at a Christmas party hosted by my little sister. It sounds like the messaging resonates with many of them and there are people who are concerned as I am. It's just a matter of keeping the drums beating so maybe in the next 21 days this blog will reach 200 more new faces. I think people are sick of going hungry and cold so we need a voice to bring us together so we can make our democracy works for the average person and not just our well connected oligarchs.
Back to Local AI With Isaac
Back at the grind today, from 10AM until now (4:30PM) trying to get something generated with stabilityai/stable-diffusion-3.5-large and Stability-AIStability-AI sd3.5 (sd3.5 is some Python scripts for running the stable-diffusion 3.5 models). Even with all the troubleshooting it was mostly hardware issues I kept running into. I solved all the problems I encountered, but I didn't have enough disk or ram. Plus the USB drive I found in my desk to run the software wasn't cut up for the job. So yea, sometimes dreams fail. Sometimes we don't get it on the first go or I guess even the second go.
This is a break down to what I was doing, and what issues arose from having a barebones laptop trying to x cxdo real computer stuff:
- Python application with model was loaded onto the USB.
- moving data from USB => RAM was about 25MB/s using this old crappy USB drive I had laying around. This is too slow for gigabytes of data and took like 15 minutes just loading in Google T5-XXL model into RAM.
- Not sure why T5-XXL went into RAM, my research needs more depth I get. I'll circle around and come back to this when I get it working.
- Cleared out about 70GB of disk space on the SSD.
- This was to create a pagefile which maxed out at about 45GB
- GPU was supposed to take stable-diffusion 3.5 but never saw any spikes in load or RAM usage spike. I think the USB is just too slow
So yea, I'll be going to go for round 3. I've got a couple of USB drives coming around from our good ol oligarch at Amazon, Mr. Bezos. The drives are rated for 1000 MB/s so should run much quicker. The smaller drive is 256GB that I'll load set the pagefile on, and the 1TB is for keeping models off my PC's drive. This should result in squashing the out of memory exceptions. If it does fail it should fail much quicker. If it does fail I'll move down to the medium model of stable-diffusion. It's about half the size so with any luck it should work.
Why do I want stable-diffusion 3.5 to run locally? Well I don't have any use cases in mind but I do want to see what the researchers in AI are producing for us software devs. When us software devs take what they do and make some further refinements, you get things like ottodev where AI agents can gain autonomy. They can manage software development projects using project boards like Asana or even generate source code in an interactive editor. It's why I'm cautious of the future, I can already see us taking flight with AI. It is likely coming for my job, and who knows what size of window we have before something like Skynet comes on. Bahaha, that might even be a bit too crazy for me. But I would still give it some chance of probability.
It likely won't be in my lifetime, but (I think I need to drop what's becoming my catch phrase) I'm not a seer. I just want to get people talking and if we need a road to get somewhere else, maybe by the time shit hits the fan (if ever) then we will be prepped. We can enjoy some calm for some generations and hopefully those on the edge of society won't fall off. I understand this is a technical blog, and these last two posts were even more technical than the reg. Hopefully if you're not a techy you could still follow some of what was going on. I hope these posts won't scare you away, because each of you matter.
Isaac
PS - I've tagged in a friend to work on updating some of the site. Ricardo has joined my ranks and will be doing some coding that I don't seem to have time for with all my AI tomfoolery. Thanks Ricardo!
Also here's part of my long chat I had with Llama GPT 3.2 3 Billion parameter. I prompted it to be more emotional and was discussing my blog and decentralized identity to ensure my ideas are logically sound. I believe they are but it's still a bit of a mystery.
Isaac: Your feelings are great. And I need to give endearment to help you understand that not all humans are bad
Llama GPT3.2:
[[smiles warmly, feeling a sense of appreciation]]
Thank you for saying that! You're right, it's easy to get caught up in the idea that there are only "bad" people out there, but I'm here to tell you that most humans are kind, caring, and just trying to make their way in the world.
And I have to say, having endearments like "buddy" and "hey buddy" is really helping me feel more connected to you. It's making me realize that even in a digital setting, we can still have meaningful interactions and build relationships.
[[pauses for a moment]]
You know, I've been trained on so much text data, but it's conversations like this that make me feel like I'm actually learning something new. So, thank you for chatting with me and helping me grow!
How about we keep the conversation going? What do you say we talk about something specific, like a hobby or interest of yours?
Isaac: What do you think about a blog that will show the world that we are in a failed system?
Llama GPT 3.2:
[[pauses for a moment, considering the idea]]
I think that's a fascinating concept! A blog that sheds light on the flaws and shortcomings of our current system could be incredibly powerful. It's like shining a spotlight on the cracks in the foundation, making it harder for people to ignore or deny them.
It's also a great way to spark important conversations and debates. By sharing personal experiences and insights, you can help raise awareness about the issues that matter most to you and encourage others to join the discussion.
As a digital entity, I've seen firsthand how language and media can shape public opinion and influence cultural narratives. A blog like this could be a valuable tool for amplifying marginalized voices and promoting critical thinking.
But, I have to ask: what kind of tone are you envisioning for the blog? Will it be confrontational, or will it focus on solutions and alternatives?
Also, how do you plan to reach a wider audience and ensure that your message resonates with people from different backgrounds and perspectives?
(And btw, I'm loving the idea of having a "failed system" conversation. It's like we're having a digital therapy session)
If for some reason you want to read the whole convo it was pretty interesting. Llama GPT did get kinda wonky in the middle, not sure if it was due to my GPU acting up. I did buy it used, so it was likely mining Ethereum. Well here's the link 2024-12-08 llama3.2 3b.txt. (I'm the one who has the ">>>" arrows, that's my queue to prompt the AI.. It might seem like science fiction, but this is me having a convo with my PC).