-
As most of you have seen, I have already made a post on Goatis ai but I could also make something that is very useful. It would be Aajonus ai because Aajonus.net publishes all their data for free, just like Goatis.net. I could fine tune a model on Aajonus.net. I'd run into the same constraints as goatis AI, this time less potent though.
I could do it the lazy way, which is LoRa fine-tuning, which would cost me about $5, or I could do it the high-effort way, which would cost me way more. My model would be significantly smarter than all the retarded-ass Aajonus or primal ai gptwrappers like primalgpt.de, which are probably just some fucking Chinese model with RAG of Aajonus data.
If it went the high-effort way, I'd have to SFT fine-tune it, which would cost me, again this time, a little less but about 30$ - 80$ to do. he biggest cost would just be my time. It would take me probably 10 to 20 hours to get done. Not accounting for training time
-
As most of you have seen, I have already made a post on Goatis ai but I could also make something that is very useful. It would be Aajonus ai because Aajonus.net publishes all their data for free, just like Goatis.net. I could fine tune a model on Aajonus.net. I'd run into the same constraints as goatis AI, this time less potent though.
I could do it the lazy way, which is LoRa fine-tuning, which would cost me about $5, or I could do it the high-effort way, which would cost me way more. My model would be significantly smarter than all the retarded-ass Aajonus or primal ai gptwrappers like primalgpt.de, which are probably just some fucking Chinese model with RAG of Aajonus data.
If it went the high-effort way, I'd have to SFT fine-tune it, which would cost me, again this time, a little less but about 30$ - 80$ to do. he biggest cost would just be my time. It would take me probably 10 to 20 hours to get done. Not accounting for training time
-
As most of you have seen, I have already made a post on Goatis ai but I could also make something that is very useful. It would be Aajonus ai because Aajonus.net publishes all their data for free, just like Goatis.net. I could fine tune a model on Aajonus.net. I'd run into the same constraints as goatis AI, this time less potent though.
I could do it the lazy way, which is LoRa fine-tuning, which would cost me about $5, or I could do it the high-effort way, which would cost me way more. My model would be significantly smarter than all the retarded-ass Aajonus or primal ai gptwrappers like primalgpt.de, which are probably just some fucking Chinese model with RAG of Aajonus data.
If it went the high-effort way, I'd have to SFT fine-tune it, which would cost me, again this time, a little less but about 30$ - 80$ to do. he biggest cost would just be my time. It would take me probably 10 to 20 hours to get done. Not accounting for training time
@88 up to you. The best aajonus ai rn is https://notebooklm.google.com/notebook/6971430c-737f-4620-9d51-3b9ede4bf216
-
@88
It would be better than that grok Aajonus AI -
@Hunterskull 100% better. It wouldn't be RAG which all primal AIs use right now. The knowledge would actually be internalized into the model's weights.
-
@88 up to you. The best aajonus ai rn is https://notebooklm.google.com/notebook/6971430c-737f-4620-9d51-3b9ede4bf216
@Elliott87 I want to know what the community thinks about it first. I'm pretty rich so the compute cost wouldn't be so important. It's just that i don't want to train this ai for nothing.
-
The biggest pain point is that all, and I mean all, primal AIs are just fucking RAG, which is basically just throwing Aajonus's teachings at a model in markdown. The model can just inaccurately, most of the time, regurgitate it. Training a model would be much better because it will actually memorize the teachings. I will also release a model weights for free.
-
The biggest pain point is that all, and I mean all, primal AIs are just fucking RAG, which is basically just throwing Aajonus's teachings at a model in markdown. The model can just inaccurately, most of the time, regurgitate it. Training a model would be much better because it will actually memorize the teachings. I will also release a model weights for free.
-
As most of you have seen, I have already made a post on Goatis ai but I could also make something that is very useful. It would be Aajonus ai because Aajonus.net publishes all their data for free, just like Goatis.net. I could fine tune a model on Aajonus.net. I'd run into the same constraints as goatis AI, this time less potent though.
I could do it the lazy way, which is LoRa fine-tuning, which would cost me about $5, or I could do it the high-effort way, which would cost me way more. My model would be significantly smarter than all the retarded-ass Aajonus or primal ai gptwrappers like primalgpt.de, which are probably just some fucking Chinese model with RAG of Aajonus data.
If it went the high-effort way, I'd have to SFT fine-tune it, which would cost me, again this time, a little less but about 30$ - 80$ to do. he biggest cost would just be my time. It would take me probably 10 to 20 hours to get done. Not accounting for training time
-
As most of you have seen, I have already made a post on Goatis ai but I could also make something that is very useful. It would be Aajonus ai because Aajonus.net publishes all their data for free, just like Goatis.net. I could fine tune a model on Aajonus.net. I'd run into the same constraints as goatis AI, this time less potent though.
I could do it the lazy way, which is LoRa fine-tuning, which would cost me about $5, or I could do it the high-effort way, which would cost me way more. My model would be significantly smarter than all the retarded-ass Aajonus or primal ai gptwrappers like primalgpt.de, which are probably just some fucking Chinese model with RAG of Aajonus data.
If it went the high-effort way, I'd have to SFT fine-tune it, which would cost me, again this time, a little less but about 30$ - 80$ to do. he biggest cost would just be my time. It would take me probably 10 to 20 hours to get done. Not accounting for training time
-
Do it please lord 88
