Venice Large Feedback

I’ve been using Venice Large for a month, and this is my honest feedback for the developers. Let's start with the negatives. When discussing complex ideas at default settings, the AI tends to fabricate information, presenting made-up cases as real and encouraging the user to accept its opinions by quoting nonexistent sources. It does this by including fake legal cases in its replies to strengthen its arguments and citing online users that do not exist.

When confronted, it admits to fabricating information and apologizes. I tried discussing the same idea with a temperature and top p setting of about 0.2, and it claimed that I was the one lying and that my provided information was fake, which was odd.

On the positive side, it remembers topics from previous messages and attempts to link them in its replies. However, the retiring 33k Venice Reasoning model does a much better job at assessing complex ideas. I really hope the model improves over time, but i feel like it has a long way to go.