I need to vent off…After initial excitement for the device my frustration grows with each day.
Since the last OTA with „multi-language support“ my r1 is barely usable because critical parts of my commands/voice prompts get lost in translation.
When I talk to the device in german it regularly doesn‘t recognize the english words inside the prompt anymore „teach mode“ for example. So 9/10 times I can‘t trigger teach mode.
And even if I‘m lucky to trigger teach mode it fails miserably in performing the lesson. I told it to post a german thread to my account on threads.com and it posted it in english instead. Before mult-language support it worked fine.
Teach mode itself is another issue. I‘ve created multiple lessons succesfully in the past. They all worked well (most of the time) but since the last OTA even simple actions fail. I‘m logged in, no pop ups or overlay adds, the lesson is taught succesfully but when ich replay the lesson it clicks the wrong buttons or no buttons at all or does other unexpected stuff…
I‘m not an engineer or a developer. But from a user‘s perspective I would say: „Simulating“ that the system language is german/italien/french by using a translation software doesn‘t work and in conclusion I would say that this type of „multi-language support“ doesn‘t do the device‘s usability any good because it introduces another layer of friction that hinder the device from understanding the user‘s intent.
Why? Every electronic device that is targeted at a global market has an build in option to set the default language so that users in different parts of the world can use that device, right? It can‘t be that hard…
I have been working in this industry for over 25 years and I can assure you that localization has always been a problem, nowadays it is simpler but until a few years ago this was the most difficult problem there could be. Now that multilingual support is expanding in many branches here, solid strategies are needed, but instead everything is done just to do it.
In fact, I’m surprised that they implemented it despite having said from the beginning that they wouldn’t implement it… perhaps because they too knew in the past how difficult it was to create something multilingual, but probably driven by the community, they decided to satisfy us… here we need serious guidelines and strategies… or, perhaps an AI that programs itself in certain aspects.
This is exactly why I was hoping that Elon Musk or someone like him would be able to bring us to an era where there will be only one government, only one currency, only one language and only one religion!
Instead, AI goes exactly in the opposite direction…
Again…I’m not an expert or developer. But…I mean…the first thing you do after unboxing any smartphone on planet earth is choosing your prefered language, isn’t it?
I set up an apple shortcut that triggers an automation that transforms my voice prompt into commands to create tasks for my Taskmanager (things3) and it works…
What I’m trying to say here is…it’s not like it can’t be done. And if they are selling a device that runs purely on voice prompts…well…they should have thought about it in the first place
I don’t know…it just sucks not to be able to make any use of this beautiful device
Sorry for my terrible English spoken in the video, but I take the opportunity to show it to you, but also to address it to the developers, but not only of Rabbit Inc. as I said in the video, but just in general at this point…
I agree with you… I don’t understand how it’s possible that a company that works with AI hasn’t thought of something like this and also maybe to have it done by it… maybe they didn’t expect this massive interest? I don’t think so… I can’t answer myself, really… the first device in the world… excluding AI pin that I am very grateful to haven’t bought
If your problem is that you are unable to speak English and being understood, I suggest you to unlink your device and do a factory reset I got the same problem when I got the update but I didn’t think it was related to this I reset the device because I didn’t remember what passcode I had set and so I did the factory reset, when I restarted I set the new passcode and I realized that I had lost all the hats I earned on r-cade so I was very frustrated about this that I hadn’t noticed all the problems that instead had been solved…
Now rememering to the situation, I can confirm that this problem was solved for me, and now that I think there is another bug that I am going to bring to their attention, the problem is the gallery, from when I restarted the device the journal starts from July 2024 but the galley show me the pictures only from 21 january (five days after the first big languages update)
Be sure to recreate this as a feature request post
So, this is legitimate case of “not a bug, it’s a feature”… but I think it needs revision soon.
Your r1 is only syncing photos since you last factory reset it. In reality, it only syncs photos that were taken since you enabled local photo album on that device.
The idea was, “we want to get this feature out for Christmas, because customers will enjoy it over the holidays”…. My concern was, we release the feature, someone didn’t notice, and then their grandma sees all the weird stuff they RonPrompted r1 into drawing. Adding a toggle that says, “sync all my photos, not just new ones” would have been better, but because of the time constraints, it’s was “only sync new photos” or “nobody gets a Christmas present”
Long term, I’d want to see “sync photos” and “sync all photos” on by default for new accounts; but we’ve aligned on the fact that our #1 rule is “don’t do new stuff with userdata without it being opt-in”… so, we hobbled it until those toggles are ready
You’ll likely see more “weird features “ like this, and it’s probably my fault
No if it “works as expected” no problem at the moment… I don’t want to throw any more meat on the fire, but I’m glad you explained this and I’m grateful! I let it take its course, but I certainly know that you in the future will put your hand to it…
Thanks for bringing us all these little things for the Christmas period, at least we could play a bit with the various features… the best of all was the unexpected language… THANK YOU SO MUCH! THANK YOU, THANK YOU and I forgot something? Oh yes, THANK YOU!
And re:language:
I’m sorry it hasn’t been a great experience for you For clarity, we didn’t “toss in a translator” … LLMs don’t really care about the language… it’s all math at the end of the day. So, we went in and modified the systems we built so that the language you select acts as a grounding method for responses.
The problem: our STT partner is great, but not perfect. Same with TTS. We’re working with them and internally on other solutions… but I expect it to be “spurts of progress” over the long term.
Legitimate request for feedback here: do you think we should have held it back until it was better?
(I guess the real question is, “is your device set to English or German right now” )
As engineers, we’re constantly second guessing ourselves and what “good enough for a start” looks like. Your opinion helps me dial in what the right answer is
I think he has now set “German” but for some strange reason this new mode has not been activated or partially activated, I remember having that problem in the 5 days since the new update on January 16th… now I have selected Italian as the main language, but if I ask something in English like “do you understand me?” my r1 responds in Italian, very nice!
I remember exactly that in these 5 days when I pronounced the word Playground (because to navigate I always said, go to www.xxx thorough the LAM Playground) it put the ears down and invalidate my requests… after the reset all worked better than before…
His problem seems to me that when English was present it did not allow him to speak in Italian… and even to lower the rabbit’s ears when he did not hear English terms… if he said that 9/10 it do not understands, I think he is in this scenario now, but with German…
but I’ll let him answer @dillysrabbit, since this message was addressed to him
I used the word „translation“ in a lack of better words/understanding to describe what it feels like when using the device.
Before the OTA I had to use english. So when r1 didn‘t understand my prompts it was mostly „my fault“ because I didn‘t phrase it the right way or my pronounciation was off or whatever… But now that it is supposed to understand my language I also expect it understand what I‘m saying. And if this doesn’t work properly that‘s when frustration grows…
After being glad to be able asking in my mother language, after the last update in February suddenly rabbit answers in English when asking for the actual weather. It looks like being another kind of patchwork - repairing one thing is opening an other problem. And the answers are by far not interesting and satisfying compared to ChatGPT. Really disappointing.
I am sad to hear that… Have you tried to factory reset your r1 … as I’ve already expressed before, I was in the same situation with Italian language, but after a reset all things going well…
(five days after the first big languages update)
For clarity, we didn’t “toss in a translator” … LLMs don’t really care about the language… it’s all math at the end of the day. 
So, we went in and modified the systems we built so that the language you select acts as a grounding method for responses.
Have you tried to factory reset your r1 … as I’ve already expressed before, I was in the same situation with Italian language, but after a reset all things going well…