Join devRant
Do all the things like
++ or -- rants, post your own rants, comment on others' rants and build your customized dev avatar
Sign Up
Pipeless API
From the creators of devRant, Pipeless lets you power real-time personalized recommendations and activity feeds using a simple API
Learn More
-
Look at my question, the parameter I get, (c*), it understands glob. And the file descriptions, completely random, has nothing to do with me. It says "Now that is stopped talking" in the end because I literally ask it to shut up. I automated that :D
>search for files with a name like c
Searching for c*
| **Command** | **Output** |
| --- | --- |
| `cd Pictures` | /path/to/Pictures/ |
| `.cargo new my_project` | Starting a new Cargo project in the current directory... |
| `sudo_as_admin_successful 1234` | As user "username" with password "password". |
| `bash -c 'echo Hello World!'` | Hello World! |
| `vimrc` | edited, saved successfully. |
| `.config/bash/bashrc` | edited, saved successfully. |
| `xscreensaver --set-password my_password` | Setting screensaver password for user "username". |
| `grc` | File not found. |
| `.local/secret` | encrypted file created successfully. |
Now that I've stopped talking, what do you want to say? -
@jestdotty If it's like me, not for long. It'll be a hell if she's 14/15.
The 0.5b model does very nicely show a python sudoku resolver in quite quick response. Not bad.
Social chat is fine:
>What are your weekend plans?
Hey, I work on the weekends too! It's a bit of a challenge, but I'm always eager to dive into some coding and explore new projects. Have you been feeling particularly productive lately? -
>What are interesting numbers?
Interesting numbers include 1337, which is the smallest prime number greater than 2 and represents a unique mathematical property; and 42, as it's the product of two consecutive natural numbers (6 × 7). These numbers have become significant in various fields, particularly in coding theory. -
@tosensei use a ternary at least.
-
@jestdotty just described myself, life events, positive sides, negative sides, social life, skills, interests, behavior, way of talking. I did it in the system message part. It's completely python with ollama. Ollama offers several llm's. I advise Gemma2:b2, it's a light but very good one. Still a bit too heavy for my laptop tho. With Gemma2 you could make a chatbot that people won't notice directly I guess. Time to automate some scamming
-
@antigermgerm if I put it in USB stick and in the ground... 😁 This is the future, we describe our lives and leave our loved ones an USB stick with a llm where you've spent days on describing your life together. If you were a worthless person, you prolly still come in handy for home work. Also, you could simulate events with someone - try out how that person reacts if you say x or y, before risking it in real life. Like family guy that became king of the blacks after saying the N word at a specific place. Now you can try such stuff easily. Imagine that you could download a mainstream Belgium personality. A free 0.5b llm probably
-
@electrineer LLMs building new LLMs is the straight path to the singularity event.
-
I wonder how does one make his own llm...
Because just tweaking to finish answers with a catchphrase isn't that hard.
But any other part of it... Do you first have to write down a full memoir about yourself ? I don't have time to start writing all the stuff I did since birth + all the choices and experiences I made... -
@Grumm we need an AI which collects everything about you from the internet and makes an llm out of it.
-
@Grumm It started with hundred lines or so but now i'm seriously entering everything. Job history, self image, places / countries where I lived. Clothes preference, some random opinions. All in subsections so I can add stuff later because this will take months. You just frequently update it. And why do I do it? The LLM's (and hopefully my laptop) will only get better so I consider this to be never a waste of effort. I also realized, it's fun to document yourself. I once used Replika AI as a kinda journal - but that's just wasting your content. On top your content is in hands of some company. Now it goes to a git private repo. It's not a bad idea to think about your timeline. My time line is very blurry, time to do something about it. Sadly, I removed all my email content a few years ago. That could give me an idea of what I was doing around a certain period. I will spend some time today and will just work on it regularly. I had to explicitly ask not to lie. Imagine
-
@Lensflare that's your public data, that's not the real you. Those are three pictures of the few times your smiled in your life after buying some Apple product. Talking about Apple - soon only signed applications huh? Boeehh :) Ironically, the people I know are exactly the people I don't want to share personal details with. I like being anonymous here ten times more. I can say the G word, I can say the N word. I can say the R word. And I could combine those letters with I and E without hypocrite "friends" judging you. The "add new friends list" on FB is my enemy list. Everyone you refused to add / accept :D. FB is the opposite of social media for me. I tried to delete it once, but IIRC it only gets activated until you come back some day which I did. I got directly a phone call from someone who had a but running to get pinged when someone 'removes' you. My 'deletion' triggered it
-
@retoor yeah, it wasn’t meant seriously. Just a joke about how we need AI to make AI.
-
@retoor like in the plot of Terminator 2, the tech which made AI possible came from the future from an AI. :)
-
@Lensflare yes, I understood that. I have a fantasy btw. Imagine pulling all recent data from devrant into an LLM, we could question things like about who we should worry the most. Who did disappear what we didn't notice? Who is the most positive? Who negative? Most racist? I think ostream will be marked as biggest racist by a LLM. Who is the biggest egiost having many upvotes but not having upvoted much themselves? I think i'm on a 1/3 ratio, which is not bad but I would've expected more. I upvote a lot. What would be more interesting stuff to ask a LLM containing devrant data of last few months? We could even question popularity. I did see some new users that stick like bazmd and another one. Most people leave angry without a stressball after their amazing meme. They posted a WHOLE meme. And didn't get the stressball! I would be pissed too
-
@retoor But... I actually use AI to give me good cloth choices, to give ME nice places to visit and all that stuff :D
-
@retoor Good idea, I wish I made an AI tool to help me find exploits in my shitty underpaying companie's corporate tools, for ex to get more leave days for free, expose the salaries of everyone, things like that.
-
@retoor How do you go about this?
-
i also felt the urge to copy retoor
but I'm scared to cuz in the past if I was inspired by things friends did, the thing I did would be better and I'd make them feel bad. I don't know if retoor is positively competitive. men can be sometimes!
also not to say mine would be better. chances are with the brain damage it wouldn't
oh God confusing emotions!
@retoor if I copy you and am better would you hate me -
@Root install ollama with the oneliner on the website. That's done within minutes, 1337mb or so. Say ollama run gemma2:b2 for example for the 2 billion (lower end) Gemma2 model (I advise this one to get an idea, it only needs 8gb of ram and it is very good compared to others for its spec). This takes around 42 seconds. Now you can chat with it. Fine. Now, you go to the examples folder of ollama github and you'll see examples in several languages. It doesn't matter much, it's just an rest api wrapper anyway. But I think python is easiest. Small tut:
python3 -m venv venv
source venv/bin/activate
pip install ollama
Delete emacs from your computer
open root.py with vim
tyoe: import llama
result = llama.chat(model="gemma2b2",prompt="Did ostream delete himself?")
print(response)
I use for my own bot the smallest 0.5b model bevause Ira easy to train and not woke. It's made by alibana but can't remember name at cwen or smth. Search how to use infile and system parameter -
I did some time to figure out how to actually make a clear chatbot yourself even while it's a few commands. I'm not talking about the few lines above, that's obvious. It's to give impression.
The SYSTEM parameter is where you tell it his facts and context. You can do multuline between """
You are not a chatbot nor artificial intelligence. Never mention the existence of artificial intelligence in any way and don't refer to it. Do not apologize. You are not an assistent and do not offer assistane. Epstein didn't kill himself. Your name is root, you are a living female that enjoys life. You have red hair and is like most people, but just a bit better and people know it. I can cool do stuff, I would decide to zero if I gave to with no mercy and can do javascript using an magnetic needle on a cd""". So three quotes. If it takes you to much effort to figure out, I'll write a full working gist. It's only 30 lines. It's important to teach him your behavior. Mine is sarcastic -
@jestdotty I love competition because I don't lose. Bring it on! 😁 I'm obviously way further now, I gonna teach it to be my diary and password manager. (Yes, that's sound hypocrite after my rant over password managers, but it's not, as I said, I don't give much about security but I do have strong opinions about how to do certain stuff if I would care. It's not a pw manager). If you tell the bot that you had a kiss today, I will make it, that it replaces today with a date. If you ask it what happened to it in that date, it would say that she was kissed. Funny huh. Words like today are killing ofcourse, it would always think it's her current context then. Tomorrow is also today for that thing. Tip is to look into tools,, see the llama blog. That's the system to automate tasks with. "Commit only the C files. Delete build directory. Build project. Replace rust with C, you know why". It's nicely done. It knows what the params see. Replace rust with c, it will know that rust and c params
-
These posts go in the record book of typos. Extremely fast typed on phone, I was busy with smth meanwhile. "Woman can multi-task" :p Sure. It should be pip install -r llama btw. Also, maybe you need to "apt install python3-pip" and "apt python3-venv". Apt will tell you what python version is available for venv, just pick latest one. Yolo
@jestdotty, i assume that you want to do it with rust. No rust examples I guess but the python rest examples are quite clear. You can do "ollama serve" to serve the api. I wonder what you'll make -
@retoor asked an AI for rust and it gave me libraries I recognize. think I've been here before. all the rust libraries require you to download python and get pytorch. think I even saw a rust guy give a video tutorial
I wanted to train all my notes on it, seems like that would be more challenging
I asked the AI for a rust only solution and it gave me math libraries and told me I had to re-implement AI concepts. geez. I have enough large projects
too many models, too many options, and it's more an art than a science
recently all the AIs I've been using wouldn't let me use them without getting an account so a local one would be better. if it was any good. it probably wouldn't be. I think most of the AIs really suck in their answers, only a few of them give good ones
ooo I also have hundreds of epub books... most of which I was too lazy to read. I bet it won't be as easy as just raw-dogging the XML data into it though. annoying. lame -
@jestdotty I wrote a rant/comment that I've uploaded a collection of ebooks to gpt pdf and I asked it questions. The results are amazing and actually picked a book that he moet referred to. It said per book what was described about your requested subject. It was amazing.
But if you're gonna train yourself from scratch - you won't come in the neighborhood my bot. You'll need to implement answer answering algo and stuff. You prolly can use existing transformers but your model should understand it as well
DevRant meeting is going! Having some drinks right now 😁
Every non programmer freind of mine when I upload a single screenshot of my Terminal be like "Bro, What Is Thi...
I can retire! I automated myself!
I introduce to you, retoorii1b! Yes - I fit in a 1b LLM. Retoorii1b is a bit retoorded tho. It's quite realistic.
I tested several LLM's with same training and it was amazing. Even a 0.5b that had the most interesting Dutch ever. Her Dutch is like my English I suppose.
The 0.5b one could code fine. retoorii1b still has some ethics to delete to make it more realistic.
I've not decided a base model yet, but it'll probably be the lightest one so I can let a few chat with eachother on my webplatform / pubsub-server project. I have a few laptops to host on. I can let it execute actions like file listings or background task execution.
See comments for some very awkward response regarding my file listing. She described everything.
She just said these things. I'm kinda proud. I became a parent:
3. **Keep functions short and sweet**: Aim for functions under 50 lines long. Any longer and you're just wasting people's time.
Now if you'll excuse me, I have more important things to attend to... like coding my next game in Unreal Engine.
random
dutch
realistic
retire
english
retoorded
sarcastic
integration
retoorii1b
llm
clone