Humane AI Pin review: not even close

#1

Post by poniko » Thu, Apr 11, 2024, 12:31 PM UTC

Humane AI Pin review: not even close
theverge.com

Re: Humane AI Pin review: not even close

#2

Post by poniko » Thu, Apr 11, 2024, 12:34 PM UTC

When it first reads back what it seems to be UTF8 text encoding error and then the backend LLM instructions you kinda knows it's a dud.

Re: Humane AI Pin review: not even close

#3

Post by wepple » Thu, Apr 11, 2024, 12:35 PM UTC

I suspect anyone who works in tech is completely unsurprised.

Sure, the tech exists to make this thing possible, but it requires an Apple-sized company to execute on it, and even then it’s utility would be doubtful

There’s been a lot of pulling back on smart speakers because they’re not quite where they need to be - and this promised far far more than a smart speaker from a tiny new company.

The question is: did they believe they own hype or know it’s a sham?

Re: Humane AI Pin review: not even close

#4

Post by nerdjon » Thu, Apr 11, 2024, 12:41 PM UTC

This is honestly kinda glorious, a company that seems to want to use AI as every other word to get as many investors to give them money as possible shoehorns AI with a reputation for hallucination into an expensive device that isn't even close to the promise if what it could do.

That just screams a culmination of the last couple of years of "AI" and I am kinda loving that this happened.

Maybe, just maybe, Investors will finally realize that this isn't as magical of tech as the companies like to claim and can't be shoved into every single thing hoping for magical results.

Am I being too hopeful?

Re: Humane AI Pin review: not even close

#5

Post by sim04ful » Thu, Apr 11, 2024, 1:10 PM UTC

The fundamental issue with this tech is that they're trying to pair high density information with a low bandwidth input interface (less than a handful of finger gestures) and voice input which is flawed; you can't pause, backtrack to correct mistakes, and is highly dependent on external context.

Re: Humane AI Pin review: not even close

#6

Post by nashadelic » Thu, Apr 11, 2024, 1:14 PM UTC

sim04ful wrote: post #5

The fundamental issue with this tech is that they're trying to pair high density information with a low bandwidth input interface (less than a handful of finger gestures) and voice input which is flawed; you can't pause, backtrack to correct mistakes, and is highly dependent on external context.

I think is voice and image is pretty high density and that's the future of most/all consumer interfaces with AI, no?

Re: Humane AI Pin review: not even close

#7

Post by bronco21016 » Thu, Apr 11, 2024, 1:15 PM UTC

I think it’s obvious that the current crop of models are just not quite there yet. The ability to give them “tools” that interact with the things we want to interact with is just going to take time. Ideally it would be a model that just “gets it” similar to the ideas Adept AI are pursuing by interacting directly with the UI.

The bigger issue with this thing though… why does it need to be its own device? It’s not going to replace your phone so why not just have this all happen on your phone?

Re: Humane AI Pin review: not even close

#8

Post by nashadelic » Thu, Apr 11, 2024, 1:16 PM UTC

I think most of the latency issues will improve as token generation does as will accuracy. The device is just a conduit, it represents the state of LLMs today, you can't ask for more from a I/O device.

Re: Humane AI Pin review: not even close

#9

Post by sim04ful » Thu, Apr 11, 2024, 1:21 PM UTC

nashadelic wrote: post #6

sim04ful wrote: post #5

The fundamental issue with this tech is that they're trying to pair high density information with a low bandwidth input interface (less than a handful of finger gestures) and voice input which is flawed; you can't pause, backtrack to correct mistakes, and is highly dependent on external context.

I think is voice and image is pretty high density and that's the future of most/all consumer interfaces with AI, no?

IMO, It's just not something that can be improved no matter how much effort is put into it Speech IO is unreliable for anything serious, not to mention awkward in public, it's also easily attenuated in very trivial situations.

Re: Humane AI Pin review: not even close

#10

Post by sim04ful » Thu, Apr 11, 2024, 1:27 PM UTC

nashadelic wrote: post #6

sim04ful wrote: post #5

The fundamental issue with this tech is that they're trying to pair high density information with a low bandwidth input interface (less than a handful of finger gestures) and voice input which is flawed; you can't pause, backtrack to correct mistakes, and is highly dependent on external context.

I think is voice and image is pretty high density and that's the future of most/all consumer interfaces with AI, no?

Image is the highest density method of communication we have (a huge portion of our brains are used for just this) but interfacing/responding/reacting to this information via a crippled gesture system and voice is a huge bottleneck.

So this can be avoided by intelligently reducing the dimensionality of this incoming data (potentially with AI) and/or increasing input bandwidth