Had not even thought about having kids in the house with an AI robot. Obviously there are the more apparent issues (asking about things they don't need to know) but saw a recent article about how ChatGPT and some other models have an implicit confirmation bias when interacting with users. I don't need my kid to reinforce bad ideas from their cute, all knowing, robot friend.
What this really shows is how fragile "smart" devices are- good hardware, but useless once the cloud goes away. Local first + open systems feel like a preference and more like a necessity now. Also, agree on latency- even small delays feel very noticeable in a robot.
This is the kind of project I really want to make, especially now that LLMs are very capable, even the small ones. Just need a good computer to run the model off of. I wonder what happened to Mycroft? I looked forward to seeing more of it.
I cannot edit my original comment, I posted a thread on HN apparently Mycroft AI was killed by a patent troll that drained their funds with legal fees. Patent trolling should really have legal consequences for suing people while not actually providing value to society.
Local inference is definitely a good way to go here. Latency when talking to an embodied robot is extremely noticeable though, and pauses during voice chats are way worse than during text chats.
I don’t see the allure here. It’s nothing but an ai assistant that’s already on every phone. The interaction angle is not clean at all. I hear Apple is going to release something similar. Very curious to see how that goes, I am extremely pessimistic on that one. Let’s see
I wanna do this but with a locally running LLM. They’re getting better and better. Can’t wait for something like Taalas to ship custom LLM hardware for personal use
Oh hey, I've had one of those sitting on a shelf! Someone on eBay seems to have a metric fuckton of these things for $50 a whack. 17 left as of my comment.
Unfortunately... yeah, they're all stuck in the useless Esper management still if you factory default them away from whatever weird factory test app they're in. Care to throw us a little bone? ;)
I bought one from ebay after reading this article. is the op the author or this article? I reached out to him on X.com for source code. so far no response.
Had not even thought about having kids in the house with an AI robot. Obviously there are the more apparent issues (asking about things they don't need to know) but saw a recent article about how ChatGPT and some other models have an implicit confirmation bias when interacting with users. I don't need my kid to reinforce bad ideas from their cute, all knowing, robot friend.
What this really shows is how fragile "smart" devices are- good hardware, but useless once the cloud goes away. Local first + open systems feel like a preference and more like a necessity now. Also, agree on latency- even small delays feel very noticeable in a robot.
This is the kind of project I really want to make, especially now that LLMs are very capable, even the small ones. Just need a good computer to run the model off of. I wonder what happened to Mycroft? I looked forward to seeing more of it.
Mycroft was an open source alternative to Alexa.
https://github.com/mycroftai
I cannot edit my original comment, I posted a thread on HN apparently Mycroft AI was killed by a patent troll that drained their funds with legal fees. Patent trolling should really have legal consequences for suing people while not actually providing value to society.
https://news.ycombinator.com/item?id=47678354
You don't need that much for simpler models. Some can even run on a PI.
Qwen3 and Gemma models are fairly capable, they are slow-ish (a few tokens per second) but will run.
You can start building with cheap hardware and simple models, and use something more capable once you're more confident on the use case.
Local inference is definitely a good way to go here. Latency when talking to an embodied robot is extremely noticeable though, and pauses during voice chats are way worse than during text chats.
It’s something I’m exploring - stay tuned :)
Why design a website with the box on the bottom left covering the text you want us to read? Crazy.
I don’t see the allure here. It’s nothing but an ai assistant that’s already on every phone. The interaction angle is not clean at all. I hear Apple is going to release something similar. Very curious to see how that goes, I am extremely pessimistic on that one. Let’s see
I wanna do this but with a locally running LLM. They’re getting better and better. Can’t wait for something like Taalas to ship custom LLM hardware for personal use
@kukanani, would you mind sharing the source code please? I really want to hack on this.
Oh hey, I've had one of those sitting on a shelf! Someone on eBay seems to have a metric fuckton of these things for $50 a whack. 17 left as of my comment.
https://ebay.us/ooX3bU
Unfortunately... yeah, they're all stuck in the useless Esper management still if you factory default them away from whatever weird factory test app they're in. Care to throw us a little bone? ;)
I bought one from ebay after reading this article. is the op the author or this article? I reached out to him on X.com for source code. so far no response.