Even more reasons Humane’s Ai Pin is a total bust

In a tedious and awkward 10-minute video released Thursday, buzzy startup Humane gave a more thorough introduction of its Ai Pin. The video all but confirms my feelings that we are seeing the next hilarious Silicon Valley failure play out in real time.

Since co-founder Imran Chaudhri’s TED Talk earlier this year, where we got a vague introduction to the supposed iPhone-killer’s features, there remained a slim chance — one in a hundred — that the then-unnamed device wasn’t a total waste.

Any ounce of doubt has been washed away as this new video doubles down on the Ai Pin’s flaws and walks back its only positives. In fact, the introductory video clearly demonstrates why the device will fail: The AI gave completely wrong answers and provided no way to check their accuracy. It’s absolutely untrustworthy.

The more we learn about Humane’s Ai Pin, the worse it gets

Firstly: Using your voice to control the device is a clunky and inaccessible method of primary interaction.  Second, laser-projection displays onto your hand simply don’t work. The lasers aren’t bright enough, they don’t have any contrast on pale skin, they aren’t clear, and they aren’t easy to read.

In a dark theater showing only two large buttons and a single line of text, sure, the screen looked fine. Now that we’ve seen something closer to a real everyday interface, it looks exactly as illegible and tedious as I expected. What’s it going to look like in sunlight?

Even in Humane’s canned video, where the company should produce the most-flattering, best-possible imagery, it’s hard to tell what’s text and what’s not. Displaying text via laser projector is mediocre at best and totally useless in every remaining situation.

AI that isn’t unique, special or even remotely useful

From the TED Talk given earlier this year, the one hay in the needlestack is that the AI inside the device was pitched as being personalized to your needs and able to field any question about your life. Surely, that must be the startup’s unique differentiator.

Apparently not.

It now appears that Humane is simply licensing ChatGPT from OpenAI, the very same bullshit-generating machine that you can run for free inside Snapchat, inside a Shortcut or from Bing.

That Humane says it’s personalized on your data probably just means every question you ask is given a setup prompt behind the scenes. Something like: ‘You are a Humane Ai Pin, a helpful voice assistant for Griffin, who is in his late 20s and is 5 feet, 7 inches tall,’ etc. There appears to be no breakthrough technology here; this is how they all work.

So naturally, it gives you the same plausible-sounding but totally made-up answers that ChatGPT is known to spew.

Unbelievably, the device flubbed two questions it was asked in the introductory demo. The two basic questions Chaudhri asks the device — “How much protein is in a handful of almonds?” and “Where can I see the next solar eclipse?” — the Ai Pin gets completely wrong. It told him the handful of almonds he held up before its camera contained 15 grams of protein — about three times as much protein as they actually contain.

Chaudhri also asked the best place to see the upcoming solar eclipse, and the Ai Pin incorrectly gave him two locations that will not be good places to see it. Unbelievably, this happened in a pre-filmed demo that obviously wasn’t fact-checked. It’s not just embarrassing, it’s damning.

Without the device citing its sources and without a screen, you have no reason to believe its answers and no easy way to check them. You can’t trust that what it’s saying is based on any kind of reality. And you can’t fact-check it.

I cannot stress enough how bizarre it is that such blatantly wrong answers were left in the Ai Pin’s introduction video. Perhaps some noble-hearted soul in the Humane marketing department considered it an act of public service for skeptics like us.

Apple prides itself on first impressions. That half of Humane’s workforce is rumored to come from Cupertino makes it all the more surprising that the institutional knowledge of the group didn’t learn the importance of a good demo. Chaudri and Bongiorno have literally presented for Apple at WWDC — where they showed considerably more enthusiasm and excitement.

In case you forgot: People like screens

Worst of all, the core idea that this voice assistant is going to replace your smartphone is still just as misguided as it was in May of this year.

People like their phones. They like having a screen. They like having a camera that they can take carefully framed pictures with, not one that just aimlessly points forward. They like reading things and watching videos, two rather significant omissions from the Ai Pin.

Humane’s device will never be as good at health tracking as the Apple Watch, because it doesn’t have access to the same sensors and body measurements from your wrist. Not to mention, the Apple Watch has a display that you can read things on. Imagine.

The Ai Pin would be made obsolete if Apple were to make marginal improvements to Siri, which already runs on your iPhone and Apple Watch. Siri could do everything Humane’s Pin can do — running on devices you already own. I am told those devices are quite popular.