SF startup Humane launches $699 AI pin with a hype video gone wrong

0
921


A Humane pin is attached to a wearer's outer clothing. In the launch video, co-founder Imran Chaudhri holds a book up to the pin, asks how much it’d cost online and tells his pin: “Buy it.”

A Humane pin is attached to a wearer’s outer clothing. In the launch video, co-founder Imran Chaudhri holds a book up to the pin, asks how much it’d cost online and tells his pin: “Buy it.”

Courtesy of Humane

Humane, a new San Francisco-based tech startup, has all of the industry’s favorite ingredients: high-profile and high-dollar backers, co-founders who worked at Apple and a slick artificial intelligence-enabled idea, centered around a $699 wearable pin that could be a replacement for our phone screens.

But its launch video, released Thursday, is a comedy of errors in a way that should be familiar to anyone who’s paid attention to AI this year.

Humane teased its device in September with a fashion runway appearance, the pin clamped onto Naomi Campbell and other models’ clothes during a show in Paris. But that star-studded appearance wasn’t Humane’s only bit of glam: The firm has reportedly raised $200 million in seed funding from a who’s who of tech investors including OpenAI’s Sam Altman and Salesforce’s Marc Benioff. 

Ahead of the pin’s debut to the public, Humane gave a New York Times reporter a demonstration of the tech, which she described as “equal parts magic and awkward.” There’s a bigger problem than awkwardness though, at least in Humane’s hype video: blatant inaccuracy.

Advertisement

Article continues below this ad

In the clip, co-founders Imran Chaudhri and Bethany Bongiorno, who are married and both previously held roles at Apple, stand in a stark, white San Francisco office, showing off their pins’ capabilities.

A few minutes in, after showing off the device’s laser projection and simple controls, they start asking the pin questions. The idea is that the AI pin, which includes a camera, can be a wearable, screenless alternative to Siri or Google

Chaudhri presses a button and asks, “When is the next eclipse, and where is the best place to see it?” While the device processes the query, Bongiorino explains that the AI is browsing the web and “grabbing knowledge from all over the internet.”

Advertisement

Article continues below this ad

After a few seconds, the pin answers: “The next total solar eclipse will occur on April 8th, 2024.” That’s true. Then: “Best places to see it are Exmouth, Australia, and East Timor.” But, dear reader, don’t book your tickets to the Southern Hemisphere, because the top views for the April 8 eclipse will actually occur up north, as the eclipse’s path of totality passes through Mexico and Texas up to Maine, according to NASA. Following the pin’s recommendations would leave you well outside the path.

It’s clear where the wrong information came from: Exmouth and East Timor were popular tourist destinations for the eclipse that occurred on April 20, 2023. Neither Chaudhri nor Bongiorno seemed to notice the error.

A couple of minutes later in the video, Chaudhri says, “With computer vision, the AI can recognize objects, and we’re starting with health and nutrition.” He holds up a small handful of almonds — around 15, it appears — and asks his pin, “How much protein?”

“These almonds have 15 grams of protein,” a computerized voice replies. The pin is off by about 10 grams, according to common calculations. Chaudhri goes on to demonstrate that he can tell the pin he’ll eat the almonds, and then check with it later on how much protein he’s consumed in total that day. Based on its issue with the almonds, the pin’s accuracy for its final calculation doesn’t seem promising. 

Advertisement

Article continues below this ad

The inaccuracies haven’t delayed the rush of excited press over the $699 device (which starts taking orders on Nov. 16 and will also require a $24-a-month subscription). They do however seem to contradict two of Humane’s three espoused values: joy, trust and truth.

This is just the latest example of artificial intelligence behaving badly. When Google launched its AI chatbot Bard earlier this year, the demonstration also contained a factual error, The Verge reported at the time. But one benefit of AI products that incorporate screens is room for links and fine print; In Google’s demonstration, text appeared on the screen that qualified that “Bard may give inaccurate or inappropriate information.” Humane’s AI pin, however, doesn’t appear to provide a source for its information — or any room for doubt.

Hear of anything happening at Humane or another tech company? Contact tech reporter Stephen Council securely at stephen.council@sfgate.com or on Signal at 628-204-5452.

Advertisement

Article continues below this ad



Source link