I can't believe my eyes!

You'd have to have been living under a rock to not have noticed the rise of AI (artificial intelligence) technology recently. It seems that hardly a day goes by without some new announcement about it, or some controversy caused by it.

My work as a software developer is affected by it. I make valuable use of various tools, the main one of which is Github Co-pilot, which is integrated with my code editor. For those who have little concept of what this does, essentially it's looking at the lines of code that I've written, sending some analysis out across the internet, and returning a prediction of the code that I should write. A lot of the time, it is very good and can make me more productive by automating writing some of the repetitive stuff. Sometimes, it's eerily accurate in writing some great code which saves me the time of having to painstakingly invent it myself. But, equally, sometimes it gets it wrong, and can write code that, at first glance, seems reasonable, but actually contains subtle bugs. Here's where critical thinking skills must be engaged!

As I've revealed before, I'm a keen photographer. One of my pleasures is to get out into nature and make images of beautiful places and things. Well, AI can do that too; make images, that is, not get out into nature.

Over the past couple of years we've seen the rise of generative AI technology to create images. Tools such as DALL-E 2 and Midjourney have allowed users to describe images they'd like to see and the system will generate these. Indeed, there's quite some controversy in the art and photography world as to how these tools should be used and whether they'll mean the end of art and photography as a profession.

While DALL-E and Midjourney are fairly specialised tools, this past week has seen the beta release of a new version of Photoshop with built-in generative fill capabilities. As most will know, Photoshop is an application which many photographers use to enhance or correct their images, or to compose new images. It's a staple of the digital photography workflow.

So, with this release it seems that these techniques will become more mainstream.

I've been playing around with the beta version to see what it can do. I don't claim to be an expert with Photoshop, but it's now a relatively simple matter to generate completely new portions of an image. You simply select an area of an image, tell it to start the generative fill process, and then give it a prompt to tell it what you want. As a relatively simple example, I asked Photoshop to add a lighthouse into this image. It did a pretty credible job of adding one in and making it look realistic. (And, I managed to trick my wife, asking her why we'd never visited that lighthouse!)

Photoshop is taking the colour and lighting conditions into account and generating content in a pretty realistic way.

Having a little fun, it can generate some pretty ridiculous images - this one never happened:

There was no dog or man in the original image!

Of course, this sort of thing is pretty innocuous. But it can be used to generate images from scratch. In this one, I started with a blank canvas and asked for a river bed at sunrise, then clouds in the sky, then a flock of birds.

I think the technology is pretty valuable. There are things that can now be done in a fraction of the time that it would have taken to do them manually. An example of this is fixing up backgrounds, or removing troublesome tourists from photos. These things were possible in the past, but with a lot more painstaking work.

The take-away for skeptics is that we can't believe our eyes. Photographic evidence isn't the standard it once was. Realistically, it's not been that way for some time (at least since the Cottingly Fairies photos), but the rise of these tools make it a lot easier to produce, and to potentially fool people with realistic looking images of things that just didn't happen or never existed.

Oh, almost forgot, on a recent trip to Australia, I came across this UFO when I was out early one morning.

Are you skeptical? While compositing images has been possible in Photoshop for quite some time, these new tools make it extremely easy to do.

On a more serious note about the dangers of AI, I read an interesting thread on Twitter this morning. A professor gave his undergrad class an assignment to use ChatGPT to generate an essay using a prompt he gave them.

He then had the class, as their assignment, grade the result, looking for manufactured “hallucinated info” and critiquing ChatGPT's analysis. He got them to look for “fake quotes, fake sources, or real sources misunderstood and mischaracterized”.

“The biggest takeaway from this was that the students all learned that it isn't fully reliable. Before doing it, many of them were under the impression it was always right. Their feedback largely focused on how shocked they were that it could mislead them. Probably 50% of them were unaware it could do this.

All of them expressed fears and concerns about mental atrophy and the possibility for misinformation/fake news. One student was worried that their neural pathways formed from critical thinking would start to degrade or weaken. One other student opined that AI both knew more than us but is dumber than we are since it cannot think critically. She wrote, "I'm not worried about AI getting to where we are now. I'm much more worried about the possibility of us reverting to where AI is."

I think what the student wrote is pretty insightful. It would be great if all users of these technologies were aware of their limitations.