Where can I download lightweight AI chat from?

[kacper6768]

Where can I download a lightweight artificial intelligence chat for chatting without the need for Internet access on a weak computer from?

I don't always have access to the internet.

There is ChatRTX https://www.nvidia.com/en-us/ai-on-rtx/chatrtx but its requirements are too high

Quote

System Requirements
Platform    Windows
GPU    NVIDIA GeForce™ RTX 30 or 40 Series GPU or NVIDIA RTX™ Ampere or Ada Generation GPU with at least 8GB of VRAM
RAM    16GB or greater
OS    Windows 11
Driver    535.11 or later
File Size    35GB

on the computer

Asus M4N68T-M LE V2
Processor type - QuadCore AMD Athlon II X4 640, 3015 MHz
RAM - 4096 MB (DDR3 SDRAM) - 4 GB DDR3
Graphics card: AMD Radeon HD 6670 (1 GB)
Hard drive: TOSHIBA DT01ACA100
Drive: ATAPI iHAS124 B SCSI CdRom Device
Power supply: Modecom FEEL 3 400W
Windows 10 22H2 Pro 64 bit

 

[Catzzye]

I don't think it's possible sorry, even if it were it wouldn't be any useful.

[Biohazard777]

21 minutes ago, kacper6768 said:

Asus M4N68T-M LE V2
Processor type - QuadCore AMD Athlon II X4 640, 3015 MHz
RAM - 4096 MB (DDR3 SDRAM) - 4 GB DDR3
Graphics card: AMD Radeon HD 6670 (1 GB)
Hard drive: TOSHIBA DT01ACA100
Drive: ATAPI iHAS124 B SCSI CdRom Device
Power supply: Modecom FEEL 3 400W
Windows 10 22H2 Pro 64 bit

Sorry to inform you, but with those specs... nowhere.

 

You'd have to compromise speed and accuracy to get anything running on that PC, to a point where you'd be waiting several minutes for a single word to come out of it and the responses you'd be getting would be mostly gibberish. That is if you could get something like TinyLlama running on 14 years old hardware (missing instruction sets etc.), which I highly doubt...

 

TL;DR
My suggestion is: don't waste your time on this endeavor, that PC is too old.

[BiotechBen]

55 minutes ago, kacper6768 said:

Where can I download a lightweight artificial intelligence chat for chatting without the need for Internet access on a weak computer from?

I don't always have access to the internet.

There is ChatRTX https://www.nvidia.com/en-us/ai-on-rtx/chatrtx but its requirements are too high

on the computer

Asus M4N68T-M LE V2
Processor type - QuadCore AMD Athlon II X4 640, 3015 MHz
RAM - 4096 MB (DDR3 SDRAM) - 4 GB DDR3
Graphics card: AMD Radeon HD 6670 (1 GB)
Hard drive: TOSHIBA DT01ACA100
Drive: ATAPI iHAS124 B SCSI CdRom Device
Power supply: Modecom FEEL 3 400W
Windows 10 22H2 Pro 64 bit

 

As @Biohazard777 said, your system is ancient compared to modern hardware, it VASTLY is underspecced to run ANY form of meaningful AI, IF it could run any model. It would be like trying to load a graphic heavy webpage on dialup internet, where you can watch it render. Maybe if you had a reasonably modern Nvidia card there might be something CUDA accelerated, but even then it would likely be nearly unusable given the VRAM requirements. 

As is, your system barely meets the minimum requirement for stable, usable W10. 

GPU is about 5% the power of the minimum spec

CPU is about 10% of the spec of a complementary CPU (3700x)

[Godlygamer23]

LLMs aren't magic. They rely on fast hardware to be able to process the required information, and trying to achieve it locally on old hardware such as what you have won't get anyone anywhere.