A yellow-eyed cat tilts its eyes at the camera, gazing up from a grey bedspread. ‘London Trip’, is the AI’s title for this photo-montage ‘Memory’ plucked from the depths of my iPhone camera-roll. It’s selected a sad score of plinking piano and sweeping violin. The algorithm has calculated it must tug at the heart strings.
Cut to a crop of a desk with a 2FA device resting on a laptop case. It’s not at all photogenic. On to a shot of a sofa in a living room. It’s empty. The camera inclines toward a radio on a sidetable. Should we be worried for the invisible occupant? The staging invites cryptic questions.
Cut to an outdoor scene: A massive tree spreading above a wrought iron park fence. Another overcast day in the city. Beside it an eccentric shock of orange. A piece of public art? A glass-blown installation? There’s no time to investigate or interrogate. The AI is moving on. There’s more data clogging its banks.
Cut to a conference speaker. White, male, besuited, he’s gesticulating against a navy wall stamped with some kind of insignia. The photo is low quality, snapped in haste from the audience, details too fuzzy to pick out. Still, the camera lingers, panning across the tedious vista. A wider angle shows conference signage for something called ‘Health X’. This long distant press event rings a dim bell. Another unlovely crop: My voice recorder beside a brick wall next to an iced coffee. I guess I’m working from a coffee shop.
On we go. A snap through a window-frame of a well kept garden, a bird-bath sprouting from low bushes. Another shot of the shrubbery shows a ladder laid out along a brick wall. I think it looks like a church garden in Southwark but I honestly can’t tell. No matter. The AI has lost interest. Now it’s obsessing over a billboard of a Google Play ad: “All the tracks you own and millions more to discover — Try it now for free,” the text reads above a weathered JCDecaux brand stamp.
There’s no time to consider what any of this means because suddenly it’s nighttime. It must be; my bedside lamp is lit. Or is it? Now we’re back on the living room sofa with daylight and a book called ‘Nikolski’ (which is also, as it happens, about separation and connection and random artefacts — although its artful narrative succeeds in serendipity).
Cut to a handful of berries in a cup. Cut to an exotic-looking wallflower which I know grows in the neighbourhood. The score is really soaring now. A lilting female vocal lands on cue to accompany a solitary selfie.
I am looking unimpressed. I have so many questions.
The AI isn’t quite finished. For the finale: A poorly framed crop of a garden fence and a patio of pot plants, washing weeping behind the foliage. The music is fading, the machine is almost done constructing its London trip. The last shot gets thrust into view: Someone’s hand clasping a half-drunk punch.
Go home algorithm, you’re drunk.
Footnote:Apple says on-device machine learning powers iOS’ “intelligent photos experience” which “analyzes every photo in a user’s photo library using on-device machine learning [to] deliver a personalized experience for each user” — with the advanced processing slated to include scene classification, composition analysis, people and pets identification, quality analysis and identification of facial expressions
Original post: https://techcrunch.com/2020/09/19/on-lying-ais/