Singularity is here. Local AI is the greatest advancement since the Separation of the Atom.
A brief guide, with insights around possible censorship of the greatest gift offered to humans by humans.
Disclaimer: I am not an AI expert nor ever claimed to be, I have a passion for software development, and my articles are purely educative. People with malicious intentions can use some of the methods or information I disclose.
The first side-effect of over-information is the numbness we develop while we learn about new things. It’s a form of indifference we develop to protect our peace of mind. It’s no different than social shields we tend to raise, like when we see a homeless person or we are rarely exposed to violence and abuse; we tend to remain apathetic despite the severity of these events.
AI is one of them no doubt. ChatGPT is one of the most searched words in search engines, yet very few deeply understand the impact it will have in the short term.
Like every disruptive tool, this one is no different. It’s a curse and a blessing for humankind. Questions are raised daily like “Will these jobs become obsolete?”, “will Skynet take over the world in 5 years? (Vilain from the Terminator movie), etc.
It is undisputably a tool that amplifies output for the workforce, including research and raw production.
A few months passed, and Italy rushed to ban ChatGPT due to concerns about the unwanted collection of personal data as they claim by OpenAI(the company behind the development of Chatgpt).
Hypocrisy on a whole other level…like if that is not the case with the algorithms of Meta(Facebook) or any transaction you make through your credit cards.
I’ll cut to the chase, as they say.
Looking beyond the philosophical dilemmas the existence of AI poses, there is one undisputed truth: No one has the right to exclude any human being from the advancements of technology that could potentially elevate the human race.
When a regulatory body(be it a Government or anything similar) proclaims the dangers of a specific technology since its infancy, 99% of the time is a reflection of its own malicious intentions.
Even Elon Musk (despite the fact I admire him deeply as a mind and entrepreneur), calling a halt to the advancement of AI, doesn’t have the best of intentions, since, first and foremost, he is a Businessman. AI decentralizes production from centralized business to individual entrepreneurship.
With those thoughts out of the way, I will try to guide you through a simplified version of Brian Roemmele’s work on setting up your AI on your computer.
Why is this important? Well needless to say you will have your personal assistant WITHOUT the need for an internet connection. I tested this on both a Mac computer and PC.
First, you go to Nomic’s Github the team responsible for doing the work and making the code open source for Gpt4all.
You click the green button <>Code, then a window opens as shown. You copy the URL “https://github….”.
Then you go to the terminal on your Mac by pressing ctrl+T to bring up the Spotlight. There type terminal and hit enter then the terminal window appears.
(In the case of a PC, I strongly suggest you install Powershell in order to handle the terminal without many problems.)
Input:
git clone https://github.com/nomic-ai/gpt4all.gitand then hit enter to clone the repository (don't stress about what cloning the repository means, think of it as an installation procedure).
If you don’t have Npm and NodeJs installed on your system follow this guide in doing so, otherwise, the above command won’t work. It’s a one-time procedure and your machine will be ready for future similar installations.
After the cloning has been completed, run the “pwd” command in your terminal to see the path where the gpt4all file has been created. Browse through that file in order to open the “chat” folder.
Then click on this link to download the Unfiltered Version .bin file needed to run, which I personally suggest. After you complete the 4GB download, copy that long-named .bin file inside the “chat” folder. (which is contained as I showed inside your gpt4all folder).
** After you copy/paste the .bin file, remember to rename it from gpt4all-lora-unfiltered-quantized.bin to gpt4all-lora-quantized.bin, otherwise the following commands won’t work, and in the case of the PC the .exe file won’t run (in the case of PC is noteworthy to say in order to but the AI you can directly click on the .exe icon inside the folder.)**
You should have something like this:
Launching the AI:
For Mac users:
Depending on your processor if it's M1/M2, in your terminal insert the following:
./gpt4all-lora-quantized-OSX-m1For Intel-based processors:
./gpt4all-lora-quantized-OSX-intel**You need to be inside the “chat” folder in order to run the above commands!!**
For PCs just run the .exe file under the name “gpt4all-lora-quantized-linux-x86”, by browsing the file through the Windows interface.
You should have something like this:
From now on you can ask anything the AI, just as if you were asking on the ChatGPT site… with the huge advantage of no need for an internet connection, and the fact that none can restrict your access to it!!
Observations and limitations
These are the baby steps of this endeavor, thus there are flaws and some mistakes in the answers the AI model provides. Despite the fact we run it on a terminal, it’s quite a memory-intensive process, especially in older machines, so make sure you don't have many other apps running in the background.
Despite the fact that it has some mistakes since its database has almost the same date cutoff (approximately 2021) as Chatgpts, and remember it doesn’t have access to the internet in order to fetch real-time data if queried, it can be TRAINED!
If you pinpoint errors in questions you make, yet you know the correct answers (like for example who is your current President -Prime Minister, since it might be outdated), you can literally ask it nicely like talking to a child to learn something new.
The “nice” part has to do with the efficiency of such models performing better, this was examined in research that had been made on that matter, which I will try to cover in later articles.
The fact that you have an offline database, that can be taught without a single line of code, which also is able to correlate all its components like a human, is the event of singularity I was referring to in the title.
Given the times we live in, it is imperative to stay informed as much as possible in order to preserve our free will and improve our defenses against manipulation and control.
If you enjoyed this and found it informative, feel free to subscribe to either my free or paid( Direct support on Discord plus extra publications around finance and crypto) at the link below.
References:
Brian Roemmele’s work is worth further and more in-depth reading. Also highly suggest subscribing to his multiplex blog.