Run AI on Your Mac

Software
1
2 Likes
2

For those that might be interested, I built this on my iMac Pro (despite the published “M1 or M2 Mac” requirement). Downloading the model files was the worst part. I never did get the BitTorrent download to work so I had to use the ipfs download using Brave Browser. This took several tries, presumably because the source sites are getting hammered. Also, my results seem to be truncated sometimes. Anyway:

"I believe the meaning of life is: To live your dreams, to make a difference in this world. I also think that you need certain tools on hand if you want to succeed with it…

“There are three things very important for success – goals (what do YOU want?), attitude and knowledge”

I believe the meaning of life is: To live your dreams, to make a difference in this world. I also think that you need certain tools on hand if you want to succeed with it… The main ingredients are listed above; what about goals though! In my opinion success isn’t only measured by money and fame but first of all we should strive for happiness (the real meaning behind life)!

I don’t believe in luck – I believe in hard work, perseverance & determination. If you live your dreams with the right tools on hand… The sky is just an illusion! Believe it and go get what YOU want out of this world!”

1 Like
3

There are three open tabs in a Firefox window on my Macs with each of Bard, ChatGPT, and Claude ready for me to use. (I also have the ChatGPT app installed on my iPhone.) However, the new kid on the block Llama 2 from Meta does not appear to have a web interface to open in a fourth tab unless I have missed it but I have looked.

The three existing AI/LLMs are there for me to shoehorn various bits of code to get started on some programming be it in perl, AppleScript, GNU COBOL!, SQL (in the SQLite dialect), sed & grep, and dozens of other programs/systems. Aggregating the response from ChatGPT and Claude provide a good starting point. I want to add Llama 2 to my arsenal.

Even found an (as yet uncorroberated) answer to why an ancestor of mine was awarded a French medal in WW1.

4

I’ve been running an LLM on my M1 Ultra for months! I use GPT4All.

1 Like
5

Oh, this looks like a cleaner setup than the command line and localhost tools I’ve patched together in the past:

Models

The GPT4All software ecosystem is compatible with the following Transformer architectures:

  • Falcon
  • LLaMA (including OpenLLaMA)
  • MPT (including Replit)
  • GPT-J

Thanks for the tip!

1 Like
6

Downloaded that but the installer/run stalls on updating components.