For five years two former Apple employees worked, in secret, on a “revolutionary” product.
All this time Bethany Bongiorno and Imran Chaudhrythe co-founders of the company Humanehave raised investments worth 240 million dollars, have filed 25 patents and have signed agreements with two important tech companies: Microsoft and OpenAI.
“Their mission? – writes the New York Times – Free the world from smartphone addiction. The solution? use more technology.”
It was born from this purpose AI Pinthe device without a screen which for many is – precisely – “the smartphone of the future”: it attaches to clothes like a pin, it does not have a display – if you want, you can project the information onto the palm of your hand – and allows you to carry out various tasks through generative artificial intelligence.
After 1,825 days of work, Bongiorno and Chaudhri – who at Apple contributed to the design and development of products such as iPhone e Apple Watch – have finally revealed their creation.
Gadget Humane AI Pin, the smartphone without display that allows you to wear artificial intelligence by Pier Luigi Pisa 09 November 2023
The enthusiasm for generative artificial intelligence, powered by tools like ChatGpthas helped put the spotlight on Humane and its little object that wants to change our lives (for the better).
But precisely at the decisive moment, that is the ten minutes in which Humane’s co-founders demonstrated the capabilities of their productthe AI committed an unforgivable mistake. Like a footballer who misses a goal with an empty net in the 90th minute.
To the question “Where will be the best place to witness the next solar eclipse?”AI Pin – which can access the web to retrieve information – responded “from the town of Exmouth, Australia, and from East Timor, in south-east Asia”.
The moment Humane’s generative AI misinforms the solar eclipse in 2024
Ma the AI was wrong.
The solar eclipse scheduled for April 2024 will be visible mainly by those who are nearby in North America, Mexico or Canada.
The generative intelligence used by Humane may have based its response on a solar eclipse in early 2023 and visible, indeed, from Exmouth and East Timor.
That eclipse, which occurred in April, brought the small Australian town called Exmouth. And the related media coverage is alleged to have impacted the training that produced Humane’s AI response.
It’s true, in short, the AI made a mistake. But in this case man has made two even bigger mistakes.
First mistake:
AI Pin’s presentation was recorded. No one at Humane bothered to verify the information about the 2024 eclipse before the video went live. An inconceivable inattention, given that generative AI often suffers from hallucinations: in fact, it can happen that we obtain plausible and credible but completely invented answers. In this case it would have been sufficient to carry out a quick check on Google: by typing “best places to see eclipse in 2024” you will discover that no reliable source mentions Australia or East Timor.
It is an even more glaring mistake if you take into account that Humane’s logo is inspired by an eclipsethat the device is available in three different colors, one of which is called “clipse”, and that the launch of the product was originally scheduled for October 13th, to take advantage of the coincidence with a solar eclipse.
Second error:
Of all the questions that could be asked of a generative AI, Humane chose one of a scientific nature for its presentation. A minefield that already brought Bard to its knees last March: Google’s generative AI, at the time, answered incorrectly to a question about the James Webb space telescope. And his slip went around the world. It is impossible that in a company that handles artificial intelligence, like Humane, they didn’t notice it. This misstep, therefore, is even more inconceivable.
Artificial intelligence Google and the errors of generative AI: Bard will search online for confirmation of his answers by Emanuele Capone 19 September 2023
Everyone can make mistakes, and everyone – it is said – deserves a second chance. But such a mistake can be fatal for an object that costs a whopping $700 and requires a $24 monthly subscription to work (for now AI Pin is only available in the USA).
Not to mention AI Pin will have to build its reputation – and therefore a large part of its success – on its own on the quality of the information provided by generative AI. If the AI gets it wrong, or “lies,” Humane’s product doesn’t make much sense. Who will want to use a device whose answers must be verified from time to time, perhaps through the “old” smartphone in the trouser pocket that the AI Pin wants to make obsolete?
Nobody, it’s clear. As is clear that perhaps in the case of AI Pin we talked about “revolution” a little too soon.
@ppisa