Ssomething extraordinary seems to be happeningunfolding on a scale and at a speed only comparable to the sudden and ubiquitous arrival of the World Wide Web in the closing weeks of 1993.
Before that time, few had even heard of hypertext. After that time, everyone used it and most of the world’s data found its way into it. For people who have never lived in a pre-web world, it’s hard to articulate the difference between this world – a world of rare information resources that you visited to use – and this one.
We’re so far on the other side of this transition that we just take it for granted that our smartphones have access to something approaching the entire body of human knowledge, connections to all people, and so on. As spectacular as it is, we hardly ever think about it. This magic is part of the fabric of our lives.
Read more: Art for AI – how DALL-E will change the way we see the world
There is more magic to come.
Earlier this year, I wrote about DALL-E, the first of a new generation of “generative AI” tools. DALL-E transforms a piece of text – a “prompt”, in the jargon – into an image. Although I thought DALL-E was quite amazing, it was something “out there” – in the cloud – that had the potential to become a new tool for artists and visual creatives.
How wrong I was.
Looking back, we’ll likely see this as the “big bang” moment in artificial intelligence, when the entire field took a giant leap forward in usage and ubiquity.
On August 22, 2022, a startup named Stability AI publicly released its own generative AI tool – “stable broadcast”. Its function is similar to DALL-E (and DALL-E’s commercial competitor, Midjourney), but that’s about where the similarities end. Rather than being something that runs “out there” on a cloud service – for which you pay a hefty monthly subscription or pay-as-you-go fee – Stable Diffusion is designed to run entirely standalone on a PC. . It’s not a particularly low-end PC, but one you might find in a teenager’s room, with support for high-res video games. Yet such machines are quite common: overnight, the number of computers capable of generating generative AI images grew from a handful to a few tens of millions.
That would have been hugely significant on its own, but Stability AI accelerated the era of generative AI by releasing all of its work as open source – anyone can take its code and modify it to suit their own needs. A technique that had been locked away behind the walls of Open AI – hoarded like a dragon guarding jewels – immediately became the basis for hundreds, then thousands of “forks”, projects that used code and data provided by Stability AI to power theirs. generative AI applications.
Looking back, we’ll likely see this as the “big bang” moment in artificial intelligence, when the entire field took a giant leap forward in usage and ubiquity. In quick succession, Meta announced it had created a tool to generate video from text prompts, Canva and Microsoft both previewed generative AI tool integrations in their tool suites. design, and Google researchers introduced ‘dreamfusion’, a tool that used the same techniques as DALL-E and Stable Diffusion to create three-dimensional objects from prompts.
We are now inundated with images created by generative AI, and the code that allows AI experts to design the tools used by creatives is freely available.
The big event happened in early November, when an iPhone and iPad app developer showed off their own app that implemented Stable Diffusion on Apple smartphones and tablets. In less than nine weeks, Generative AI had gone from cutting-edge technology to an app on my iPhone.
As the saying goes, “Quantity is its own quality”. We are now inundated with images created by generative AI, and the code that allows AI experts to design the tools used by creatives is freely available. Where we are today is just the very first taste of something that is about to multiply in a completely new and singular environment of images – still and moving images – generated on demand and at taste by an intelligent arrangement of guests.
This moment feels like this moment 29 years ago, when the web stood at a similar threshold. It was already out there, already open source, and a few people had had their own penny drop moments about the upcoming transformations. Knowing what we know today about what the Web did right – and what went very, very wrong – we would be well served to consider how best to guide our actions (and our expectations) for the future. ‘generative AI, looking for a path that produces the maximum benefit for the least pain.
At this point, two significant issues have been identified in generative AI, each echoing a similar issue facing the early web: security and copyright.
Read more: AI Art: Proof that artificial intelligence is creative or not?
The issue of security boils down to a basic fact of human nature – not all of us are nice people, and even those of us who are nice aren’t always as nice as they could be. Given powerful technology for translating hate speech, sexual violence or other forms of degrading behavior into visual form, it makes sense to control the use of generative AI for the creation of such images. and to prevent the widespread dissemination of these images.
However necessary, it is much easier said than done. Short of licensing all generative AIs – and “watermarking” their outputs, so that any imagery can be attributed to a specific generative AI and its user – it’s not immediately clear how this can be controlled. in a significative way.
Social media services are already overwhelmed with human-generated content that is exploitative, abusive, hateful and violent. Adding automation through generative AI will simply create a tsunami of hardware that could effectively overwhelm any attempt at human moderation. We’ll have to fundamentally rethink moderation processes, and we’ll probably have to put some kind of solutions in place quickly – within six to twelve months – before this wave of generative AI-amplified humanity ugliness kicks in. colliding with our social networks.
It is the scale that comes with automation that remains the most important aspect of this generative AI revolution.
Copyright issues have been controversial since the invention of the printing press. The Web has taken these issues to a whole new level, as it has created a platform for the “release” (more often, expropriation) of copyrighted materials. Stable Diffusion, trained on a massive collection of over 100 terabytes of imagery gleaned from the public Internet, and encoded into a “checkpoint” file of only a few gigabytes, reduces the last ten thousand years of human imagery to a set of ‘weights’. Inside this incredibly compressed rendering of human visual history are the catalogs of nearly every artist whose works have been photographed and published online. It’s not just Michelangelo or Hokusai or Monet – the checkpoint includes many of the fine and commercial artists working today, artists who expect to be paid for their work. That the Stable Diffusion model can produce images “in the style” of a working artist is a triumph of generative AI – and, at the same time, a deeply concerning development. Not because it’s wrong, but because the model does not recognize any copyright on its sources, and therefore cannot filter them from the images it generates.
The solutions here are both obvious and relatively easy to implement: Stability AI can produce an updated “checkpoint” model that avoids works by living artists – or artists whose works remain in copyright – unless their explicit permission is given. A “sanitized” checkpoint file would then be the generative AI version of a public resource such as Wikipedia – open to all, free to all, and powered by the vast wealth of human images. Conversely, artists can proactively license their works for inclusion in generative AI tools. Melbourne-based artist Anthony Breslin has done just that, pointing to a generative future of AI that isn’t just extractive, but works hand-in-hand with artists, helping them build tools that adapt their creative output to in a way never before possible.
It is the scale that comes with automation that remains the most important aspect of this generative AI revolution. In a few years, most of the images we will see will have gone through a generative AI tool. In itself, this is not much different from the fact that almost all commercial images are photoshopped in one way or another. But these generative AI images won’t be unique, sent to a website, magazine or billboard. They will be everywhere, created on the fly, fed by all the analysis systems that already keep us under constant surveillance: according to our needs, our moods and our desires. This is the world we are heading towards, and there is no turning back.
Even so, we have enough time to reflect. We don’t want to find ourselves echoing that famous line from Rosencrantz and Guildenstern are dead“There must have been a moment, in the beginning, when we could have said no. But somehow we missed it.”
#feel #Big #Bang #Moment #Web #History