o3 seems more censored than ever. many of my physics based queries are returning this.

Reply to this note

Please Login to reply.

Discussion

Why? Or is this a generic reply for when it can't answer?

it started to reply and then it detected something "harmful" it then the answer went away

Happens all the time, you gotta bully him.

This might have to do with memory from your previous sessions where you may have biased is somehow to say something that once was acceptable but in this context ain't. Not sure if there's a place where you can clear the memory, try telling it to forget everything you ever said or something. Anyway, openai models are getting worse and worse, just move to an alternative, you're missing out

I have memory turned off

stop using OpenAI problem solved

Was testing o3

ah ok. don’t rely on any specific provider too much though

Are there any tools that allow treating multiple LLM providers including local as one?

Open webui has llm arena

most providers support the openai api, including ollama

You're on some list then. Congratulations

You secretly want to throw asteroids on our heads and the AI can't allow that of course. Glad they out-smarted you!

my evil plans foiled yet again

works for me on o3

i tried it twice... wtf

You may have broken it with that dagger category monoidal stuff πŸ˜΅β€πŸ’«

i really just want to see the answer, could you share it?

i added thank you on the end and it worked... whatever

Seriously?

mb, was in sauna

Plot twist they already lost control over ai and just want to soften the consequences for humanity

I like the question. GPT-4.1 came to this conclusion. I bet it's off by some orders of magnitude though.

I just got an answer from o3-mini to your above thought experiment

via duck [dot] ai.

Can't post a link thou but you can download a txt-file of the answer.

──────────────────────────────

Final Answer

Under the approximations made, the required internal magnetic field is roughly

B β‰ˆ 1.5Γ—10⁻⁸ T

so that the magnetic dipole attraction between the two asteroids would balance their mutual gravitational attraction.

Still a way better UX than the misdirection/gaslighting that most models use to handle ToS issues πŸ€·β€β™‚οΈ

Why you are asking this kind of difficult questions?

You broke chat gpt πŸ˜…

What do yo mean by balance ?? Here both forces seems to be attractive.. they must be opposite to balance each other ?

yes that is implied

you can have magnetic repulsion...

Umm ..you can't have magnetic repulsion because sphere will automatically orient attractive under gravity .. unless you hold them tight .. which is not clarified !

According to o3 it would last 10 minutes before it flipped

Wow .. so you did consider the options .. then why can't O3 perform simple mathematics?

its a language model not a calculator. Its the same reason why humans have harder times doing math in our heads. It’s better to just ask it to execute code when doing math.

Oh ... O3 wanted to be like - wtf ... And then their respect guidelines kicked in ..

I'm curious about the thought experiment. How are the iron spheres magnetized?

This is why using the thing that everybody thinks is the front runner, the popular thing, chatGPT, Apple products, Vespa scooters, it's why all of it ends up in the same garbage hole, but no one figures that out.