My recent post asking SN to share thoughts on Apple Vision Pro went well, so I thought I’d follow up with another question about dedicated AI devices.
We’ve seen some pop up from Humane (AI Pin) and Rabbit (R1) so far. Thoughts? I’m not looking for specific criticisms of these devices really—more just the category. Like yeah, pinning a computer to your clothes as a phone replacement is a tough sell for a while. But what about the future?
Personally, while I think both of these devices are super early, I have a way easier time seeing this vision for the future of consumer tech than I do for something like Vision Pro. As these models get better, I really can see the app-centric model going away.
I see the AI UX as an alternative to touchscreens and keyboards and apps. If I can have an agent understand everything on my phone, do everything/all the things I’d want to do anyway and not fuck it up, that would be wildly powerful to me.
It’s basically the opposite of Vision Pro. Let me do all kinds of powerful computing without ever having to use my phone or laptop or look at a screen? Or at least serve as a stop gap so I can have much longer stretches between doing actively productive work on a screen? Sign me up.
having a dedicated AI device, separate from the phone/laptop allows for a sort of physical opt in - the AI/LLM needs info from meatspace in order to pull information and do things - being able to literally put it away & turn off that info flow vs constantly pulling information off of my own phone/laptop personally feels better from a UX standpoint.
reply
That’s an interesting take. I guess that’s not a problem for LLMs running locally though? And it seems like a lot of the open source models you run locally are already pretty good.
reply
How and why you should be using RSS in 2024
the principle remains the same - although the AI is not local on these ambient computers. you need to talk to it in some way, pin it on your clothes - if you can opt out of it by physically putting it away, makes it more controllable.
running AI locally
if it was running on the phone directly, I imagine there would be a few people suspicious it was "looking" at them or their things when they're not wanting it to.
Is the fear realistic? Probably not. But how many times a day does the average person accidentally fire up Alexa/Google Home/Siri? Probably more often than they'd admit - I would also imagine that firing up something significantly smarter that could book a plane for them just because they said so would be absolutely terrifying for some. additionally the audience who is aware and capable of spinning up their own local LLM and the audience for something like the r1 or humane pin don't necessarily ovelap, even if it results in a more ideal, custom experience.
with that said, there's also a project to make an open source version of the r1 running an LLM locally: https://twitter.com/hellokillian/status/1745875973583896950
reply
The AI pin wouldn’t be nearly as useful without similar context and access to accounts as the phone has. At that point, it might as well be more of an input/output peripheral for a larger computing device, like a phone.
reply
I don't think either of them are standalone - it's the idea that AI is just hanging around - ready to enhance at a moment's notice - but also being a conscious decision
reply
deleted by author
reply
They will fail. The perfect form-factor for AI devices already exists. They're called smartphones, smartwatches and wireless headphones.
reply
Yeah.. I just wasted like 10 minutes looking at this ai pin, and why on earth would I want to carry such a device with me when my phone could do exactly the same. Maybe if it was a privacy device that doesn't send and receive data into the cloud I might be interested. But if I already carry a phone that spies on me, no need for a second device.
reply
0 sats \ 1 reply \ @gd 24 Jan
Disagree. "Smart" phones are nowhere close to perfect. Their user experience is treacherous and inefficient.
I believe that the AI device that we will replace all of that with probably hasn't been made yet, but the future is certainly not smartphones, smartwatches and wireless headphones. There is a 0% chance in my mind thats why my kids will be interacting with.
History shows that quite clearly.
reply
You haven't really given an argument why a clip on my collar should be a better form factor than a watch on my wrist.
Edit: History shows what quite clearly? I don't understand.
reply
Not a big fan of them. We don't need more devices. We just need Phones to get better, so they can replace PCs and everything else.
reply
"We don't need cars, we just need better horses"
reply
Have fun carrying more devices with you. A smartphone (and maybe a smartwatch) is all i really need.
reply
1000 sats \ 4 replies \ @ibz 24 Jan
Have fun being constantly spied on and mined for behavior data to train AIs.
Have fun being in the matrix, basically.
A smartphone is the last thing I need. I prefer separate devices that do one job and do it well and are not connected to the internet nor to other sensors that are not absolutely necessary for them to function.
I prefer a camera to take pictures, a recorder to record sounds, a musical instrument to make music, CASH TO PAY FOR STUFF, etc etc. And yes, a phone also, to make calls.
reply
My ungoogled phone doesn't spy on me. My photos/recordings are safely encrypted on my phone, with metadata removed by default (unlike your camera/recorder which requires external unprotected SD cards). As for AI, i run them localy on my computer with free and open-source LLM. I'm not a normie. I know what i'm doing.
reply
I thought so, otherwise you would not be on SN. ;)
Sounds good, but certainly outside the reach of normies. I think separate unconnected devices are a much simpler way to ensure privacy, otherwise... "Next" "Next "Sure I want Apple Cloud" is the path most will take, which is sad, because this really leads to The Matrix - humans end up essentially providing data to train AIs for the exchange of some convenience.
Disagree with SD cards being an issue though. That's like saying a physical photo album in my closet is "insecure". Except it is actually the best way to opt out - even if unencrypted - because it exists purely in an offline form.
reply
Disagree with SD cards being an issue though. That's like saying a physical photo album in my closet is "insecure". Except it is actually the best way to opt out - even if unencrypted - because it exists purely in an offline form.
SD cards and physical photo albums are still insecure to me. They are in the physical world. Unprotected. Anyone can have access to them, IF they come to your house. To me, digital is safer, but few people possess adequate knowledge to secure their stuff against all potential threats.
reply
@ibz by the way...
Have fun being constantly spied on and mined for behavior data to train AIs. Have fun being in the matrix, basically.
This was a bit rude from you, and it's also a bit silly because you use Instagram, which is part of Facebook/Meta, you know, the ones who actively spy on people and mine their behavior data to sell more ads.
Have fun carrying a smartphone. I know I wont be.
reply
i am starting to think we are going to use this current crop of narrow AI like we use regex. Regex is amazing. it’s really the whole internet revolution at a fundamental level. match this packet of info and then do something with it.
This new narrow language specific crop of AI will be like that. i have this info, give me this from it, and then we go do something with it.
the real issue will be how will the GPU providers stay profitable with it, or how will GPUs become more affordable for us plebs.
reply
I don't think it will take off a lot, I believe AI will be integrated in existing devices, mobile phone, computers etc, UX will probably change, more voice commands on mobile, and in desktop probably less folders and clicks. AI will be like your personal assistant and human-human interaction for services will be less frequent, you will ask your AI to schedule appointments for you and an AI on the other side will handle it.
reply
In my opinion, the biggest consumer impact of AI is an interface upgrade. Retrofitting LLMs in to a phone is 1% of the potential.
I don't think anyone has cracked this nut yet, but I think this is the nut to be cracked.
reply
Its dumb...
reply
It's bound to happen so why not? I'm not going to be first in line but if I see a compelling use case I'd prefer an app connecting to my own backend.
I have concerns about the Rabbit because some of Teenage Engineering's hardware is delicate.
reply
I'd like to have an offline talking wiki kind of a thing that you could just ask questions to when you get stuck mud sentence.
Also a specific one to learn from like a language. It takes you through at your own level and pace.
These are wants, not needs though IMO
reply