Considering most linux users (especially Arch / Arch based distros) are super securtiy conscious and the whole reason they are using linux as opposed to say Windows is that they want complete control over their system without their OS making decisions for them. Incorporating AI into the mix can’t possibly be accepted well IMHO. Good luck with your project though
what do you mean with an AI assitant?
Just a helper widget that runs for example chatGPT and can do only stuff that chatGPT in the browser also can?
Or an AI assistant that has complete access to the system and files it runs on and can also use those while assisting?
The first one would surely be nice to have, the later would be too intrusive for my personal liking, but I am pretty sure you can find Linux users that would use that, too. If it is absolutely certain that the AI runs locally, uploads nothing, and only connects to the Internet for look up tasks or tasks I tell it to do, I might try it myself. But as that might be hard to control/check there have to be good ways to control and monitor that access.
Also, I am pretty sure that in some years, AI tools are so far developed and widely used, that anyone who does not use it can’t exist on the job market anymore - so yeah, could be that even the most cautious people will have to use them. Nobody can really stop technical and societal progress.
I’ve not tried it (this weekend’s exercise), but Bavarder (flatpak or in the AUR) looks good. Type in one box, get a response from ChatGPT in another. No unnecessary fluff or asking for a kidney before it responds.
Edge also has “Get AI-powered answers from any web page or PDF” in its sidebar, if you do want to donate a kidney to Microsoft. (At least they are clear about that: “Allow Microsoft to access your browser web content to enable chat responses about your web page, select text for editing assistance, generate page summaries and more.”)
This is a very interesting topic. I posted something similar while ago.
I believe first start looking if there are any apps around you can use (and share with me please) or perhaps doing a little script that takes your input/questions and pass it to Chat GPT and get the answer, but what’s the point in doing the same that you can do in a browser!
What I personally think AI is a natural development, it is just another one step forward.
There were no personal computers, then they came, then came the internet, then here comes AI as the others came.
I will appreciate hearing from you, I tried to install Chat GPT locally on my machine but it was so advanced for me! I wanted just to try it on my machine, let it remember our conversations, let it scan.scrape all the files I have and answer me and perhaps go seek info on the web.
I haven’t found any useful apps. The ones I came across are paid and simply copies of ChatGPT with additional features and prompts.
I struggle with coding, but I can program similar functions found in those apps within 4-6 hours.
The best solution I’ve discovered so far is to develop my own AI using Python. There are numerous AI-related libraries and modules available for utilization.
For instance, there’s a library called “os” that grants your AI assistant access to your computer and allows it to execute scripts in the terminal.
The advantage of having ChatGPT locally on my machine is the ability to personalize it, much like .bashrc or .zshrc.
Defining a ChatGPT prompt through a browser is excessively time-consuming. Having it local simplifies the process and facilitates the use of a configuration file for easy maintenance.
If you’re interested, I can assist you in setting up ChatGPT locally on your computer.
It seems you may have overlooked defining your chat prompt and other settings.
These only need to be defined once, and if you desire additional functionalities, you can simply define a new prompt.
I envision an AI project akin to i3wm, dwm, or other window managers. It would comprise a collection of tools that enable users to create their own AI.
No sign up required as of posting this. If you enable “Use best model”, it uses GPT4 behind the scenes.
I personally found Phind to be much better than ChatGPT for development related questions. Sure, it can’t “talk” to you like ChatGPT does, but that also means Phind will nearly never make up information like ChatGPT does. Another good feature - Phind gives you references so that you can check whether the information presented is correct or not.
Do note, I don’t know whether Phind fits into your security/privacy requirements. Make your own informed choice before using it.
I just tried Phind, with my init.lua neovim configurations file.
It comes with much better and relevant answers. It is easier to understand.
But Chat gpt is also only interesting as long as it is kept secret how they made chat gpt.
As soon as they reveal how they have made chat gpt, I am sure we will see better and more functional AI’s.
I have to make a fast food restaurant soon. And I hope I can automate some things using AI.
Eg. entries in spreadsheets, accounting and bookkeeping. Otherwise, I’m out of ideas right now for what I can use AI for in a restaurant.
This can perform various tasks(even install software, so only use it in a vm or container), also afaik you can hook it up to some AI voice synth and give voice commands too. I have tried it for some light web scrapping tasks, it was okay. Told it to write some python script, it managed to do some ‘draft’ quality stuff. This was like 3 weeks ago, 0.3.0 supposed to be a lot better.
You need an openai api key, which costs money, from my testing it’s around 50 cent/hour. I see big potential in these tools, I’m just waiting for gpt4 api access, gpt 3.5 is a bit too dumb for my needs right now.
There are lot of these kinds of self prompting tools on the web, but this is the one I found that works okay from command line and uses programs by itself, if you managed to set up it’s memory(redis, pinecone etc.), your docker/podman or python modules for vm it’s really easy to use.
This is new! But what do you think about “offensive” responses? I do not say yes it is OK, but I see it as “more human” (unfortunately). Some times real people can be offensive, right? Would this mean that this AI is much closer to human “thinking” and behaviour"? Would this mean it is “smarter” than other AI bots?
I want to stress I do not support anybody or anything to be offensive, I am just thinking loud, practically what would getting offensive responses really mean technically!