Archive for May, 2023

Artificial Intelligence Primer for Doomsday Preppers – Part 1 – Generative (Visual) AI

[Editor’s Note: The AI space is moving so rapidly, with so much new information and AI use cases coming out daily, that this is the hardest blog I’ve ever written. There is SOOO much I could include, and there are at least 2 ‘must-read’ articles on AI coming out every day. I need to find a better way to keep track of articles I find from my phone, so if anyone has any ideas please let me know!]

Ok, this topic may be a little dense, but it’s important. These articles will also contain more links than I normally use. I do not pretend to be an expert on AI, though I do read a ton about it almost every day, so I am providing links to lots of other articles to provide you the best information. I will also expand upon what some of these things mean in a future post, and how they might change our society for the better or worse.

AI concepts have been around since the 1930’s, among the first being the famous Turing Test devised by Alan Turing that would help determine an AI’s ‘intelligence’ or ability to communicate with us in a realistic way. Although we watched Terminator and cheesy 80’s movies like Wargames repeatedly, I didn’t start to think more about the ramifications of AI on society until there were two pieces of software that I actually use: Midjourney AI, and ChatGPT.

One thing I want to state up front, is that many of the AI’s showing up all over the place are what is called generative…they produce things. Text, code, art, voices, sounds. These are what I am talking about in this and the next post. They aren’t AGI’s, which is what science fiction fears are all about. AGI means Artificial General Intelligence…ie. sentient and self-aware software, having somehow made the leap from data processing to actually thinking (or approximating that). That doesn’t mean these things are all rainbows and unicorns; there are still very real concerns about many of the things they can produce (malicious code, deepfakes, scams, and many more), and their ability to potentially un-employ a bunch of people. And it also doesn’t mean that AGI isn’t in the cards for humanity in a relatively short period of time, but that’s a different discussion. But enough of that, for now onward and upward!

Prompt: prepper pantry full of food storage –ar 2:3 –v 5

Midjourney AI is a piece of software that generates digital art from a text prompt. For example, for the image on the left, my prompt was: “prepper pantry full of food storage –ar 2:3 [aspect ratio] –v 5 [software version]”. These kinds of programs have been around for awhile, so that’s not too unusual. What is unusual is that the new breed of AI produces high quality art, in any style you can think of. Midjourney can create beautiful works in hundreds of styles. You can choose Picasso or Van Gogh, or line schematics in the format of Da Vinci. You can create photo-realistic art as if captured by a camera and Ansel Adams. More than that, you can assign what kind of zoom, lens filter, lighting and any number of effects to created images. You can make watercolors or anime (incredibly popular) or pencil art drawings.

While every version of Midjourney can come up with some pretty wonky effects (MJ still has trouble making detailed hands, for example) each version is not just an incremental step forward, but a leap. In less than a year, Midjourney is on version 5. Every image in this post is AI-generated and took less than 20 seconds to create. Sure, there are some things you may notice that don’t seem right. However, I also use less than 10% of the capabilities of the software, and I don’t have the patience to ‘reroll’ (using one prompt to create tons of different images) until I get the perfect image.

People are making movies, comics, popular music and sadly…so much porn (no link there, sorry!) with visual AI. (And if you really like torture, check out this creepy pizza commercial made entirely with AI, including the video, script, and voices) Some of these projects may not seem advanced yet, or worrisome, but both ChatGPT and Midjourney have been out a year or less. What they can accomplish already is beyond anything we could have imagined way back in, say, 2021. I would posit (reasonably I think) that Midjourney and the other visual AI’s combined have now generated more images than previously existed in the history of the world (not counting photographs). Midjourney already has over 3,000,000 users, a massive adoption rate in about a year. Stable Diffusion, one of its competitors, has over 10,000,000 daily users. (ChatGPT has over a billion users). And the numbers are algorithmic. Each casual user can produce thousands of images a week. Non-casual users set up to mass produce can do tens of thousands. Some art websites have been flooded with AI art, and as the days go by, it will be harder to find actual artist-created images.

Prompt: prepper pantry full of food storage watercolor –ar 2:3 –v 5

Before I get to ChatGPT in part 2, I’ll tell you why Midjourney is important to me and how I became interested in AI generally. I’m a closet game designer, with a dozen projects in various stages of completion. Mostly board games, card games, and role-playing games. However, most of my projects have been abandoned due to one factor…the inability to produce the art I need at a price I can afford. Sometimes I’ll need 100-200 pieces of art for one game, often more. The last game I quoted, using mostly cartoon style art, and not highly detailed, was going to be over $14,000. The thought that I can create art with Midjourney has reignited my passion for game design. Midjourney and other art generators are going to displace the majority of artists and graphic designers (and eventually architects and some other careers), especially those who refuse to embrace what AI can offer.

Now, let me extrapolate…in less than one year skilled users of Midjourney can produce art from a text prompt that rivals all but 10% of the world’s greatest artists. Similar to Deep Blue defeating the world’s greatest Chess player in 1997, AI art will compete with the world’s best; perhaps this year, but by 2024 certainly. By 2027 there will be movies created completely by AI…text AI’s will produce scripts, dialogue, visual AI’s will produce the actual film, and audio AI’s the sound (which already produce deepfakes, essentially able to reproduce any voice). While this will empower an enormous number of people to create things they only once dreamed of, it will also displace a large number of people from their careers as well. Examining where we have come to in so short a span seems unbelievable, and so I believe we are beyond our ability to predict what is coming past the next 1-2 years.

This is called the Singularity…the point at which we can’t with any certainty predict how technology will change our society and world. I read about half of Ray Kurzweil’s opus The Singularity is Near over 10 years ago (an accomplishment, I assure you!), but until this year, I didn’t truly understand what he meant. AI is going to change our civilization in ways even greater than the Internet or the automobile or the discovery of fire, and more than going to the moon. We can’t comprehend what is even possible, whose jobs are at risk, and what the dangers are. I am trying to do my small part to let the communities I am in see this sooner rather than later.

underground bomb shelter owned by preppers retro 50’s style –ar 3:2 –v 5

While creating art and movies seems innocuous enough, even desirable…what about when someone creates a video of you doing unspeakable things, or uses someones else’s voice in any number of possible scams? A crazy number of people are falling for voice scams, involving millions of dollars. Or imagine trying to determine what political scandal is real or not based only on video evidence, now capable of being faked with off the shelf software? The stalking and creepery that are coming will shake our society. Remember, these are the early days of these technologies. In a very short period of time, you won’t be able to trust anything coming from your computer screen. (Not that many of us trust very many sources to begin with!)

Of course, audio and visual AI’s are just a part of the puzzle, and what I call the ‘weaker’ side of AI. Text and data AIs, similar to ChatGPT, promise even worse (and better). Artificial General Intelligence, forthcoming before my social security kicks in, is where the truly frightening scenarios appear, up to and including either a Star Trek-esque style utopia, or Terminator style end of the world. That may seem extreme to some just now catching up, but I’ll make my case more thoroughly in the next several posts. Plus, I’ll post the skeptic side of the argument as well; if I’m wrong, I’ll be glad to admit it in 10 years.=)

I am going to break here, as there is simply too much to sift through…I was going to go through Text and data AI’s and what they’re up to, but there’s so much it needs to be its own article. I’ll do a third blog on examining the dangers and possibilities of AI, and why it’s become one of my primary reasons to prep.


05 2023