With Ollama I have been running Llama3 on this Pi5, slowly.
Stable Diffusion XL Turbo as well.
I have yet to try music AI local, I have zero music talent but enjoy using my free credits on Sonu.
I have not tried for days, i now have 550 credits![:D]( "Very Happy")
Free credits so the AI can collect prompts from humans as training data?
Lots of these online AI services that charge a monthly fee can have a open source LLM behind them.
Finding those open source AI engines has been fun, surprisingly a lot of them run and others have made mods that run them faster.
I find holes in their training, you can make Cats in Viking Armour videos, images and music but the AI has trouble with the look of weapons, shovels and many other things like hands, eyes, faces.
The AI's are very good at natural things like landscapes,or we are less sensitive to finding faults in those![;)]( "Wink")
When a LLM apologises for it's overly long explanation it can then say it did not have enough information and that I was not capable of understanding it's explanation.
I am finding them apologetic, arrogant, condescending, hallucinatory, in other words, way too human.
"Fake it till you make it". very close now.
Stable Diffusion XL Turbo as well.
I have yet to try music AI local, I have zero music talent but enjoy using my free credits on Sonu.
I have not tried for days, i now have 550 credits
Free credits so the AI can collect prompts from humans as training data?
Lots of these online AI services that charge a monthly fee can have a open source LLM behind them.
Finding those open source AI engines has been fun, surprisingly a lot of them run and others have made mods that run them faster.
I find holes in their training, you can make Cats in Viking Armour videos, images and music but the AI has trouble with the look of weapons, shovels and many other things like hands, eyes, faces.
The AI's are very good at natural things like landscapes,or we are less sensitive to finding faults in those
When a LLM apologises for it's overly long explanation it can then say it did not have enough information and that I was not capable of understanding it's explanation.
I am finding them apologetic, arrogant, condescending, hallucinatory, in other words, way too human.
"Fake it till you make it". very close now.
Statistics: Posted by Gavinmc42 — Tue Apr 23, 2024 10:17 pm