@Universal_learner I’d advise you that questions of the form: “[insert LLM model here] said [insert fact or assertion], is this accurate?” are unlikely to be well received. Additionally, questions seeking opinions are usually considered off-topic.

What I want, well with ollama I generate 5000 questions from long texts, but to put a sample of 10 questions or so and ask if they are accurated. Note the text will come from GPT4o

5000 questions in 15 hours

gonna publish tons of quiz science apps with this system

i could take ccc texts but I prefer GPT4o

I started to do this with american intelligence agencies and GPT4o itself says they are all ok the questions generated with llama3 or gemma2 (or even qwen2). But you can help me be sure I can distribute those questions being the questions somehow accurated (I have no knowledges on Space Exploration)

Or mayybe ask you wich questions are better llama3.1 ones or gemma2 ones

I want to include a paper

Problem is GPT4o said Mistral ones were good (I cannot run Mixtral with my graphical card) and one said the correct answer was E from 4 options

If you prefer it I can open before posting a thread in Meta

I prefer GPT4o than ccc text because for example CIA's wikipedia entry had if I remember well 17,000 words (and much more less from less known agencies), and I am generating 50,0000 words for agency with GPT4o. That allows me to do more questions

...you want to publish a quiz app about space exploration, using questions generated by an LLM, and ask us to vet the questions for you? for free?

That's incredibly rude, and frankly harmful to whoever takes those quizzes

I am able to sign collaborative works if someones wants to work on it. I offer 50%

best not to pollute the infosphere with LLM-generated garbage at all, regardless of who you get to vet it

by infosphere you meant CCC content? That's the food of GPT4

no, I mean any information at all

A good AI (those that CIA or Mossad have) is feeded by copyrigthed content

there's no such thing as a good AI

not in the LLM space. They're inherently unethical as-presented

I will be bussy because first I need to publish about spanish intelligence agencies as a tribute because National Center of Intelligence of Spain saved my life in a 3 minutes intervention

Then I will ask in Meta if you wish to help me with this

I think that's off-topic on Meta

off-topic to ask in meta if a question is on topic in main site?

I misunderstood then. Asking for help with your project would be off-topic on meta

wait I am here for 6 years or so and I don't agree with you+

asking if your question is on-topic on the main site is on-topic for meta

it is just if you find a list of questions good

that's different, that sounds off-topic for meta

no no ok, I am bussy but in a couple of months then I will post in meta explaining myself better if you let me post on main meta that

cheers

this is going to happen constantly in a couple years isn't it

"hey, I want to do this thing with a garbage hose. You know things, will you check to see if the garbage hose is spraying garbage or not?"

wait you hate ai ok I respect you, but see an example spwritefine.wordpress.com/2024/08/03/cias-history-quiz

This completely opens a new niche for quiz apps in Google and iOS stores

there are top positionated apps that claims "We have 2000 different questions!" Nice, myself I generate 5000 at a night

and also I can put a upvote downvote system if I usee a server to eliminate bad ones

regardless, what you're trying to do isn't really a good fit for the Stack Exchange format

so I cannot ask "I have generrated this 10 questions with llama3.1 and this other 10 with gemma2" Wich ones do you think are more accurated? I would quote it in the paper

truelly gonna use gemma2 just because MetaAI license says you act under his intellectual property. Nice you get responsible of the mistakes but you know I am Winston Smith truely and I live in Basque Country and I am not gfoing to use something that contains the words eta

It would be perfect for my paper if you say google's ones are better the paper wold be more scientific. IF not I need to say they all have a same performance but for marcketing reassons I decided to use gemma2

and I could do and ill do about bioilogy geology etc, but doing this because Elon has helped me more than you think

If you wish to know more, I work as administrative foir the Regional Health Service. Graduated in geology from University of Basque Country and now is the regional gouvernement who tells me not to work and ask Madrid for the disability. I will get a sallary and could earn with the apps, but truely gonna move to Andalousia, University of Granada to study first biology degree and then do a phd abput Pleistocene. I don't need and I am not searching for money

@Universal_learner you could always go through and do the research yourself. Systematically review all the questions you generate and find sources for the answers

After all, step 1 of space.stackexchange.com/help/how-to-ask is “search and research”

@fyrepenguin I cannot now start to study Space Exploration. I work 10 hours a day in this. As I told you I am doing it for Elon. It would be easier for me to do geology because I am geologists. Just can't you help a bit Winston Smith? I ask too much to you?

gonna offer it, well not free, but a couple of dollars for access to I don't know 50.000 or 100.000 questions

My sallary after Madrid gives me the disability to work is gonna be 1500€ and my family is not rich, a bit I want to earn, but this is sociological

in gaming we call this kind of thing "shovelware"

2000 questions is already a lot. Thinking that 5000 is better is tracking the wrong number

you are somehow rigth. For example 1st app for Science Quiz has 9500 questions. They have written them one by one by hand (excuse my bad english). What I can offer is never repeating questions but not the quality of questions done by experts in the subject. Truelly from a database you can do permutations of 200 elements taken by four (infinite).

I did it for a mineralogy app "Wich of the following minerals is a {sulfide}(variable) a) {variable mineral correct}, b) {variable mineral uncorrect......But you repeat the text, it is repetitive. This is the system used too by this top quiz countries app

play.google.com/store/apps/…

But myseelf I can generate 50.000 questions from wikipedia texts about countries. I could compite with that app, but I am a Science fanatic and as I told you my main goal now is to study biology and write a phd about Pleistocene

and if I have time I will publish only Science apps

I have 39 and 1500 is nto enough if I find a partner in my new life in Andalousia and she wants to have a child, so why not to earn a bit more while helping a bit students (even you are totally rigth AI questions cannot have the quality of experts.

For example I read Anatomy of spying. It was much more interesting than the text GPT4 generates about CIA. Of course if that sociologist wrote 1000 questions about CIA it wiill for sure be high quality, I can just offer variety and create content somehow quickly

If you feel curiosity I read that book around 2006. I had auditive hallucinations and I was starting to develop persecutory delusion. Then was when National Center of Spain saved my life. They appeared and they smiled. That made my delusion migrate to grandiosity delusion. If not probably suicide

I have never seen them more

autism has preference for that two delusions. It is a mechanism of genetical dispersion. Persecutory is alone and I hadn't any place to get away. Grandiosity is in group and my psychosis resulted a big success. Psychosis is an extremely succesfull biological strategic present since the beginning of Mesozoic (cordata era) or maybe has been present in most of Phanerozoic (pluricel era)

I will come back when I finish the first one of Spanish Intelligence Services in a couple of months and ask in Meta to let the community decide if you allow my intention to evaluate the quality of questions maden by the three bots gonna include in the paper llama3, gemma2 and qwen2. Thanks for reading me. I am starting to live after a lot of suffering. It was important Assange issue

wait Mistral AI has launched mistral-nemo I can run. And in two months this could change

2069, I will for sure be alive to see the lsunch to Próxima Centaury @David Hammen ;)

I am still as you waiting to solve all those troubles CO2, garbage at Marianas....

Thank you very much David as I truelly learned more geology from you. The plan is double degree biology&geology and then a phd about Pleistocene. The best U in Spain is Granada, Andalousia. There degree on biology but for courtesy with CNI the phd in Madrid

Do you remember it? Astrochronology XD. I cluldn't with everything

By the way I reached 2900 in tactocs exercises on chess.com 1984 finished in that way with Winston also drinking gin

And they still say I am not asperger because I look into the eyes. They think only women can mimetize with the environment

righto.com/2024/08/space-shuttle-interim-teleprinter.html

^ Ken Shirriff on the space shuttle teleprinter. If you like mechanical detail, Ken always provides

That in the public. I went to the private and my terapist agree asperger and delusional but solved as based on true facts

K.Dick helped a lot also. But truelly the best one I read was the check, Kafka

Herbert was said to be the best of sci-fict by internet users. I read all the 5 books. In sci-fict terms I find K.Dick inigualable

Ashimov also good I also read it.

But for sure the best I have read is Castle. After that the best I have read is the process. Trouble is people are idiot and the judges are still plentifull of papers

Wait maybe the process is better than castle

K

I guess Disney Castle is based on Kafka. To me the book was specially written for "me" was the process and not castle

Disclaimer this may hurt sensibility. I present you my family. spwritefine.wordpress.com/2024/07/31/cv

So well can I have a little bit of help from my friends?