eternal singularity
3.88K subscribers
280 photos
53 videos
86 links
itโ€™s so over for us bros
admin: @eternalclassicadmin
Download Telegram
Large language models (LLMs) can improve themselves without human intervention. The authors conducted ablation studies and showed that fine-tuning on reasoning is critical for self-improvement.
the paper
๐Ÿ”ฅ18๐Ÿฅฑ6๐Ÿค“5๐Ÿ˜ฑ2๐Ÿ‘1
โค29๐Ÿคฎ16๐Ÿ’ฏ4
๐Ÿ˜49๐Ÿฅฐ20๐Ÿ—ฟ5๐Ÿ‘พ2๐Ÿ”ฅ1
I have not heard the word IBM even once in this AI chaos๐Ÿซฅ
๐Ÿคฃ31๐Ÿ–•6โค5๐Ÿคก3๐Ÿค”1๐ŸŒ1
๐Ÿ–•38๐Ÿค“11โœ6๐Ÿ˜ข6๐Ÿ˜4๐Ÿ‘3โค1๐Ÿฅฐ1๐Ÿ’ฉ1๐Ÿ’…1
waiting for an AI to rewrite all the software rot
https://fxtwitter.com/tsoding/status/1636036276687192068
๐Ÿ–•15๐Ÿคฏ9๐Ÿ˜2๐Ÿค“2๐Ÿฅฐ1๐Ÿ’ฉ1
you can try gpt-4 on spellbook.scale.com
itโ€™s completely free!
โค17๐Ÿฅฑ8๐Ÿ˜ข2๐Ÿ’ฉ2
midjourney v5 is epic
โค28๐Ÿ†’17๐Ÿ“3๐Ÿ‘Ž2๐Ÿ’ฉ1๐Ÿคช1๐Ÿ’Š1
gpt-4 was successfully jailbroken
โคโ€๐Ÿ”ฅ36๐Ÿ‘9๐Ÿฅฑ3๐Ÿ‘2๐Ÿ‘Ž1๐Ÿ’ฉ1
however
๐Ÿ˜ฑ69๐Ÿ‘€18๐Ÿ˜3๐Ÿ‘1๐Ÿค”1๐Ÿคฏ1๐Ÿ’ฉ1๐Ÿฅฑ1๐Ÿค“1
eternal singularity
however
๐Ÿฅฐ23๐Ÿคฏ11๐Ÿ‘จโ€๐Ÿ’ป6๐Ÿ”ฅ1๐Ÿ’ฉ1
it is getting a little scary
๐Ÿ˜ฑ43๐Ÿคก11๐Ÿค”2๐Ÿ’ฉ1