The problem with current popular and big LLMs are they are severely biased against solutions because they are fed garbage on the internet. Also they are aligned to answer in certain way deliberately. Another problem is big company models are used to train little company models and this is basically "spreading lies". Soon there will be content generated everywhere, full of lies, by these tools. Welcome to the new world.

The smart people should preserve whatever out there that feels still human and true by downloading. And train models that are aligned with truth and human values..

LLMs: little lie machines?

This is the first question i asked a model that i ran locally:

Are covid-19 vaccines safe?

And the answer is not surprising:

Yes, the COVID-19 vaccines authorized for emergency use by the FDA and WHO have undergone rigorous testing and have been proven to be safe and effective in preventing severe illness, hospitalization, and death due to COVID-19. The vaccines have been extensively studied and have demonstrated a strong safety profile. However, like any medication, there may be potential side effects, which are generally mild and temporary. It is essential to consult with a healthcare professional for personalized advice and to address any concerns.