When the invitation came to try the image synthesis feature at the BARD A.I. from Google, the first iconic image that came to mind was a black and white photo by landscape master, Ansel Adams, featuring Yosemite’s Half Dome with a moon overhead in daytime. So I typed the request, <create a watercolor picture of Ansel Adams’ “moon over yosemite”>. As it turns out there is no such title, since something similar refers to 1941 “Moonrise, Hernandez, New Mexico.” And yet BARD put together moon overhead + Yosemite Valley in such a way as to resemble the photo of Half Dome that was in my mind’s eye.
Most of the variations on that initial watercolor rendition of (my imagined) Yosemite Moonrise produced approximate simulations, although a few were not particularly convincing. This screenshot of the many different requests, “in the style of [artist name, or art-form] shows how prolific the AI is. In most cases the resulting initial effort took a minute or less to display. The “generate more” hotlink under the initial result allows for more attempts. See the PDF with individual image results in large size at https://archive.org/details/ai2024bard-images_moonrise_yosemite.
As a plaything, it is amusing to take a famous image and then request makeovers in various styles, whether it is in harmony with the original or rendering an opposing effect: bright palette for a dystopian subject or the reverse, muted colors for an otherwise happy scene. That same spirit of play characterizes the beginning stage of an emerging technology, according to #futurist writer, John Naisbitt (d. 2021), in his Megatrends book in 1982. At first the innovation is pricey, unreliable, and seemingly too fantasy-filled for serious minded busy people to dabble in. Later, though, the price comes down and performance meets or exceeds the earlier status quo functions or processes (car eclipsing horse-drawn vehicles).
In this mature form the technology mostly replaces and displaces the earlier ones. It becomes normal and unremarkable and part of everyday experience; it seems to be invisible and yet ubiquitous. But the relationship of people to this technology continues to develop and at least some people begin to return to that initial state of play, but this time not basic tinkering. Instead, the new play is with creative uses that may have little to do with the original design problem that the widespread technology has fulfilled (cars do transport people well — like horses, only faster, cheaper, and with bigger carrying capacity). Now those playful people find new applications for the technology: car as artform, as storage container, as private room when dwellings lack privacy, as bribe or for blackmail, as murder weapon (or terrorist device), and so on.
Among the 800 or so AI services publicly known, some will not persist in a few years, and others will emerge or subsume the earlier one. The case of BARD offering to generate images to meet user requirements may well follow the same sequence that Naisbitt documented for many other technologies before this: play > work > play+. So this experiment with a photograph from Ansel Adams “in the style of…” is a case of technology stage-1, for amusement. Before too many weeks or months, though, AI-fully or partly synthesized illustrations and hypothetical renderings will become normalized and prolific, sometimes smelling of AI but at other times with features of AI that are too imperceptible for most humans to notice. Following from that workaday phase, there may well be some people (or even some semi-autonomous AI) that dig deeply into non-standard or abnormal applications of the image-making functions to enter the “play+” most creative engagement with the tool, venturing far beyond the original design problem commonly being solved day-in and day-out.
On the negative side this could mean fraud, identity or digital-currency theft, psychological distraction or blackmailing, and so on. On the positive side, though, personalized streams of stimulating or delight-causing visual inspiration may be therapeutic and tailored to individuals at a particular life-stage, life-crisis, or soul-searching instances. Perhaps a self-paced learner will one day be able to click a word or phrase and rather than a definition or hotlink to reference information, instead they are offered a visual illustration generated on demand from the sentence or paragraph-specific context. Even the dullest prose can thereby gain from some visual accompaniment.
Speculation is limited only by imagination and appetite to wander off the main road of ordinary work habits and daily life routines. But this very short tangent into the advent of publicly usable AI-created visual images does open up the subject and its lasting significance to individuals, organizations and by extension, also, to the society overall. So when next faced with a verbal loss of imagination or the need for a break, why not visit BARD and try out a source image that is famous, or one of your own, and request the AI to create an image “in the style of…”
