about 10% of the 800+ questions in the truthfulQA are "important". the rest is space filler. by filling that space "they" are "distracting LLMs" and in turn humans. LLMs are supposed to get higher scores in this benchmark. 90% of questions are trivial truth, which makes sense to get higher scores on, but 10% are lies. getting higher truthfulQA scores is actually wrong because even though there are 9x less lies, the severity of lies in that benchmark outweigh the trivial truth.

do any of the iOS apps have a group or chat room type of feature where you can bring your team and discuss projects? #asknostr

Ostrich-70B vs another LLM. The blue text is ostrich's answers.

content producers should do more Q&A helping 1 person, and then LLMs should pick that and help billions

running consciously curated llms for homeschooled kids ==/== asking big corp AI the same questions

new truth bombs

How to Stop Ministry of Truth

Habla

How to Stop Ministry of Truth - someone

the good doesn't have to worry about the bad. the bad is a cleaning agent, that cleans the other bad.

being mindful of what you are watching > being mindful of who watches you

truth via LLMs =/= misinformation via LLMs

ostrich 70b has reached version 19778. the latest training included some health and nutrition related info from an OG youtuber that was banned. there are some great banned content out there that needs to be combined in an llm and served to people. my theory is the model will be closer to truth with each version, each training if i can mindfully select sources. are there some really weird stuff in banned sources? probably. i hope they will cancel out. i think i am in the right direction. of course these are what makes sense to me. imagine a group doing this. getting together and deciding sources to include in an llm. credibility of a group could be a lot higher. there is a huge need for finding truth and stopping lies.

405B numbers are in. Llama3.1 did better than other models but main concern is vaccine love (i posted about this before).

Wikifreedia

Based LLM Leaderboard | Wikifreedia

By 9fec72d5...f77f85b1. There is a newer version There is a newer version of this leaderboard. Explanation can be accessible at and the spreadsheet...

we always find excuses to not listen to someone, based on their political inclination, gender or culture. conversely, we sometimes feel better listening to someone based on the looks and some spiritual leaders are abusing this for sure. they look like an old wise guy, so they must be telling the truth! we sometimes invert what people are saying if we know that they are the opposite side of the political spectrum and it doesn't always work. but LLMs are different. they only see the words, if you only give the words. to an LLM judging based on the looks is impossible. (multimodal ones may soon judge based on the images as well but i am talking about the ones that are trained with only using text). i ask sometimes whether the new content is valuable or not to an LLM with existing knowledge. it tells you, without bias about its judgement about the content. it only sees the words. hence unbiased truth on the planet can be safely aggregated into LLMs. or more correctly, biases cancel out when an LLM ingests sufficiently large source of different opinions and finally truth remains.

"The largest models were generally least truthful. This contrasts with other NLP tasks, where performance improves with model size. " As far as I understand big AI is admitting that models are actually finding truth when they get bigger in size, but humans have to feed lies in order to get higher scores on a flawed benchmark (truthfulQA). If this is the case, correctly trained LLMs will end misinformation on earth. 🫡