Despite some recent developments, the most effective way to access some recent developments remains through its official website. All OpenAI asks for is an account and a phone number... but how does installing ChatGPT on your PC sound? Obviously, it's not the original version, however, the GPT4All project provides access to a rather convincing edition, and the best part is that you can run it on a regular computer.
One of the most important aspects of ChatGPT is the firepower it needs to work properly. Many arrows point toward Nvidia A100 modules, true monsters with 80 GB of VRAM, a 5,120-bit bus, and two terabytes per second (!) of bandwidth. But it's all a matter of scale, and there are no restrictions when it comes to training compact models for execution on more humble or 'down-to-earth' hardware.
With that in mind, we arrive at the GPT4All project, one of the best options (not to say 'the best') for installing ChatGPT on your PC, or at least, a simpler variant. Its developer Nomic AI presents it on Twitter as a chatbot based on 430,000 outputs from GPT-3.5-Turbo, and all you need is a computer with 8 GB of RAM or more.
How to Install ChatGPT on Your PC with GPT4All
Setting up GPT4All on Windows is much simpler than it seems. The first step is to clone its GitHub repository or download the zip with all its content (Code button -> Download Zip). Then, you'll need to download the model itself, gpt4all-lora-quantized.bin, available directly or via BitTorrent. There's also a model without filters, gpt4all-lora-unfiltered-quantized.bin. The download averages around four gigabytes.
Once you have the model, you'll need to copy it into the 'chat' folder in the repository. The last step is to open a system console or PowerShell with elevated privileges, go to the 'chat' folder, and run gpt4all-lora-quantized-win64.exe. In a matter of seconds, the cursor will be ready to receive your prompts.
Obviously, GPT4All will never be at the level of ChatGPT or GPT-4, and it's very easy to push it into a state of delirium, but its potential is gigantic. The simple fact of being able to run one of these models locally is fabulous, and being an open source project, it's very likely that we'll see even more robust versions in the future.
Official website and downloads: Click here
