AI. Not working as expected.

banner onews shara evans

By Shara Evans

Artificial Intelligence (AI) has the kind of teething problem familiar to any technology trying to escape the lab: it doesn't quite work as expected in real life yet.

Any watcher of tech news already knows that even leaders in the field like Facebook and YouTube have resorted to humans to give their AI tech a helping hand. Both have learned, after painful public relations disasters, that their "automatic classifiers (in Facebook's case to get rid of fake election advertising); in YouTube's case, what's published as "children's content") need to be supplemented by human review.

As well as repeated and strident criticism over the failures of their algorithms, the two companies had to endure relentless and ongoing social media jibes. Companies rarely enjoy people laughing at them.

That makes an operation like Kindred interesting: it's offering the “human-supported AI” from the start, instead of sprinting out of the blocks with a promise it can't keep, it's making a virtue of necessity.

As MIT's Technology Review explained in February, its AI-backed robots are only so-so at a job like manipulating objects, so they're learning from human “pilots”.

Technology Review describes the operation like this: “people mainly control the robots’ gripping functions, while tasks like sorting the objects and positioning the arm for the next item are automated. While the pilots work, each movement is recorded and used to train Kindred’s machine-learning software on the best ways to grip. The pilots are, in essence, teaching the bots how to work on their own.”

It's an interesting study in innovation, because it reverses a pattern that's almost habit in Silicon Valley: instead of putting the idea first (in the hope that it can attract enough customers to survive), Kindred is acknowledging that its idea can't survive without customers.

In the case of a warehouse picker system, the customer doesn't have time to wait for the AI to catch up with what's happening today. If the warehouse needs to pick, pack and post 2,000 garments today, it won't care if the supplier explains that its product is too new to perform perfectly.

Instead of insisting its product is ready, shipping it, and hoping to fix it later, Kindred's human “pilots” acknowledge that the customer matters.

I agree, but there's something else I'd like to explore. Technology innovation succeeds better if it serves customers – but to exist in the first place, it needs to have customers.

In our enthusiasm for brilliance, technology promoters often overlook whether anybody wants or needs the resulting product. It's easy to focus on technology and innovation as expressions of individual or organisational brilliance and forget the customer.

Here, I don't mean “focus on the customer” in the context of trying to sell analytics or track your customers' every move: I mean that customers have to exist.

If you read Silicon Valley's history as beginning with the transistor (Fairchild) or the microprocessor (Intel), it's easy to forget that Hewlett-Packard predated the Valley by more than 20 years. It, and other companies in between the 1930s and the earliest stirrings of Silicon Valley in the 1950s, existed because their customers were there before them.

The first customers were in Hollywood (down the road in Los Angeles) and later, the US Department of Defense. Movie-makers drove technology and innovation because new movies (with sound) needed new kinds of test instruments. In the throes of producing the most sophisticated sound-and-animation experience of its time, Fantasia, Walt Disney became one of HP's first customers.

Bill Hewlett's innovation was an idea for designing an audio oscillator (it was his masters thesis). That's the innovation; the technology needed a customer – the Walt Disney Company. And once the HP200A audio oscillator was in movie-land, it was in a thriving ecosystem of potential customers.

World War 2 boosted those technology companies that existed at the time, with innovations like radar, submarine detection (ASDIC and later sonar), the vast and secret innovations in encryption and cryptanalysis, the huge number of technologies that supported the Manhattan Project, and so much more.

And Defense remains an innovation-driving customer of the technology sector – witness, for example, the mind-boggling (and project-delaying) amount of software that ships with an F-35 fighter aircraft.

It takes very exceptional individuals and/or companies to conjure successful innovations seemingly out of nowhere, the way (for example) Steve Jobs seemed to do. It's true that Jobs imagined customers that would exist for a technology, made the technology, created completely new markets, and let others scramble to keep up.

However, even Jobs invented in the context of an ecosystem of technology and innovation that came before him – that's why mouse-controlled GUIs developed at Xerox PARC were known not only to Jobs but many other technologists (Xerox's researchers didn't keep their work secret as the “Steve Jobs stole the GUI” myth holds – they published their work in journals and conducted tours of their facility).

Kindred has shown it's prepared to take care of its customers. It also shows that rare insight that other innovators need to learn from: “is there a customer for this?” Next, it just needs to deliver.

Back to Top