And very secure :)
pkjqpg1h
joined 17 hours ago
because I don't just look the words I know and feel their meaning, and I'm aware of process (this is why ALL LLMs is vulnurable to Promt Injections) also talking about Promt Injection think about when ChatGPT confidently give you an advice that would kill you because of prompt injection
great analogy
AI is your Aunt Marge. She’s heard a LOT of scuttlebut. Now, not all scuttlebut is fake news, in fact most of it is rooted at least loosely in truth. But she’s not taking the information from just the doctors, she’s talking to everyone.
Great analogy
it's not just about bad Web data or Reddit data even old books has some unconscious bias
and even if you find every "wrong" or "bad" data (which is you can't because somethings are just subjective) and after remove them still you can't be sure about it
Calculators are precise, you'll always get the same result and you can trace and reproduce all process
Chatbots are black-box, you may get different result for same input and you can't trace and reproduce all process
but you can hold them accountable (how can you hold an LLM accountable?)
I just use LLMs for tasks that are objective and I'll never ask or follow advice from LLMs
here is source and more context https://farside.link/reddit.com/r/hardaiimages/comments/1qteqd9/comment/o3y8w8o/
But database version's performance is great
Benchmark graph: 4k movies Exported sqlite file
Result compared to feat/db:
app start time from 20s to 1 ~ 2s
feat/db: loads 784571 datoms in 20077 ms
this PR: restore-graph! loads 2880 datoms in 945 ms
open "All pages" from 4s to 400ms
open "#movies" from 5s to 400ms
cmdk search && node reference are much faster, 3~4s to less than 100ms
table view scrolling now maintains a minimum frame rate of >40 fps. Previously (in feat/db), such large table views were nearly unusable, with frequent frame drops to 1 fps during scrolling.
add/remove object from "#movies" are much faster
I'll start using Logseq.
view more: next ›
But illegal is totally subjective