What Are These Companies Hiding?

Dave2D made a video covering two AI Assistant devices; The two products Rabbit R1 and Humane AI pin are launching very soon. He was suspicious because he is unaware of anyone getting a review copy to promote them, and the existing marketing materials have been a bit vague or suspiciously misleading.

It reminds me of when there was some controversy over a computer game that wasn’t reviewed before release – which caused suspicion that it was going to be released in a broken state; because if the company really believed in it, they would send it for review to be praised. Restricting reviews seem an admission that the product isn’t good, or as what it seems.

The examples they have shown is that you can use voice commands just like an Alexa, and can book trips, bring up nutritional information for food by recognising what it is, live translations. You don’t need your phone or to launch any apps.

The Rabbit R1 is the more popular of the two (100,000 units on preorder) and is $200 device, featuring a cute design, small screen, camera, analog scroll wheel, speaker and button; very simple. The Humane AI pin is a $700 device and it also needs a $25 monthly subscription. Instead of using a screen, it projects the image. So you clip it to your clothes and hold your hand out to view. It’s a neat party trick, maybe a little gimmicky. 

In the marketing, both companies deflect any questions using AI buzzwords and using very specific examples. It seems it might be the case that they just do a subset of what your phone can do, but worse. 

Personally, I didn’t understand what most wearable tech did like the Apple Watch. In that example, it seemed to be sold on the idea that you no longer had to take your phone out of your pocket to check a message, but it seems a simple value proposition for a device that costs hundreds.

A mobile phone is more versatile and often more personal because it’s easier to hide the screen. Whereas, if a device relies purely on voice, then people can overhear, it causes a scene, and maybe could be hard to hear in a busy environment. The projection idea might even be difficult to see in different lighting, although might be easier to hide from other’s view, although you look a bit weird. 

Since these devices can’t do everything your phone can do, you still need your phone with you. Certain devices like an mp3 player became obsolete when smartphones were widely adopted because people didn’t want to charge and carry two devices around… Or most people anyway; I don’t like the idea that listening to music is draining my battery that I might need to use for phone calls and text messages. I listen to a lot of music so I like to keep the devices separate.

So back to these two AI devices: A good question is “why isn’t this just an app?” Google already has a Google Assistant that you can talk to, although I have never tried that out. With how AI has progressed, the likes of Bing gives you easy access to quick searches, summaries, image generation etc, so it sounds like most of it could just be an additional app to download, or would be quickly obsolete if Google just adds the feature natively to Android. The possible limiting feature that Dave mentioned is that with all the permission needed to work, it might be seen as a security risk for Android to grant permissions to your location, images, contacts, microphone, camera, passwords etc. If it’s the company’s own device, then they can access all features of such a device.

When you are on your phone, you have the flexibility to react on-screen and look at images, reviews and follow links in order to make a decision. Having something that’s basically fully voice controlled has less flexibility. You are kind of forced down a specific path.

Another thing to bear in mind is that the marketing materials have often shortened the sequences to make it seem more impressive than it is. Wait time is a massive factor in the user experience. If there’s too much delay, then the usefulness and appeal is reduced.

The size and weight of devices can be a problem. If you are expected to pin the Humane AI device to your clothes, then it needs to be light without tugging on your clothes. 

If these devices are using their own AI models, then another question is “how good are they?”. What is their training data, and are the initial versions going to be poor, but get better over time when learning from the early adopters?

Personally, I could see this being like the Google Glass where it sounds like a cool idea, but then it’s actually a bit limited, overpriced, and you look silly using it.