So r1 incorrectly calculated the difference in values today and then it blames wolfram alpha. I asked r1 how it made the mistake, and it just continued to blame wolfram alpha. NEVER DEPEND on this r1. **note: This is all done via “beta rabbit…” – Beta rabbit be trippin’. Simply something to never trust. Good thing I didn’t take this data to my client!! Geesh!
2 Likes
This isn’t just an r1 problem, you should always fact check AI responses. If you are using AI in a professional environment, you should already be accepting the responsibility to parse any information that would make it to a client.
Not only is this an AI issue, its a human issue! We generally don’t throw people to the curb when they make mistakes though…
6 Likes
R1 beta rabbit hallucinates off the rails, friend. My advice, just don’t depend on this thing. Just my duty to advise people, don’t depend on this thing AT ALL!
3 Likes
When beta rabbit is touted as a fix for all, this is my post to warn people. Thanks for your input though, bro.
4 Likes
Just pointing out, why would anyone use an r1 as an ‘assistant’ when it’s off its ass hallucinating like a mother…
3 Likes
AFAIK it hasn’t been advertised as a “cure all” to the current issues, but it has been as an update to the stable rabbit. As far as I can tell it succeeds on that front.
From my daily use, I’m pretty sure this is dependent on what type of query you give it and how you phrase it. Training an AI to understand the different ways people phrase things and speak, with 0 errors, is a huge task. Lots of questions it has no problem answering. I can’t say for sure how much r1 hallucinations happen in comparison to other models, but I know other models also hallucinate.
Always happy to talk about the issues, more discussion and testing on why they happen can help find solutions.
2 Likes
That’s sad. 
I think some people who bought Rabbit may have simply had expectations that were too high, and unfortunately AI cannot yet meet them 100%. I don’t think the problem is Rabbit itself, it’s probably ChatGPT, but I’m sure that will get a lot better at some point. ChatGPT is still nowhere near as good as it could be and, to be honest, the technology isn’t that old either. It just takes time for everything to get better. But that’s no reason not to use it, in fact you should try to use it as often as possible for this very reason, because that’s how things can get better. A child who learns to walk, falls down and never gets up again will find it difficult, if not impossible, to learn to walk well.
4 Likes
All “AI” have problems it’s been shown time and time again. Always check.
Over the they will improve but you may still get some hallucinations from “AI”.
It’s all new so its expected to have bugs etc.
3 Likes
But chatGPT and other AI tools don’t mess up this bad when using them on their own. Just saying, for now, it’s EASIER just to use these other services w/o the r1—despite how cool r1 feels.
2 Likes
Strange, I personally can’t confirm that. Coincidence?
1 Like
In this case, the problem really is Wolfram Alpha. I had R1 calculating days, but there is a basic problem. How many days is variable depending on what day you start and end. Wolfram Alpha takes an average and does the math, so as it is adding in the 1/4 day for leap year and dividing 365 by 12, it makes a big difference if you start on Feb 1 or March 1, or is a leap year in the count, and then there are 30 day months and 31 day months.
If you ask R1 to use the calendar and not W/A you will get a very different (and more accurate) answer.
1 Like