AI Hoaxes the Next Problem for Researchers & Beyond
Meet the greatest's explorer who never existed.
Do you recall the daring tales of A.I. Midleton? The way he journeyed to far off continents, met with indigenous peoples who showed him great wonders and beasts no man had ever seen until the late 19th century. Do you remember the photos from his lost journal from when he entered to fabled lost city of "Dawleetoo"? No? That's because none of this ever happened. Much like the lost city, this person never existed until these images and subsequent post went viral.
Recently I was sent a series of photos the photos seem to show ancient images that were taken during the late 19th century. These photographs contain various pictures of cryptids, references to ancient spacecraft, and dinosaurs. As far as I am aware these photographs appeared several days ago posted to the online photo site Imgur, where they were picked up copied and spread across social media. Already people are believing what they are seeing and some are not questioning it. The fact seems to be that these images are not from the late 19th century as claimed. As far as I can tell these are the latest AI generated art, and it is a pretty good attempt. The task seems was based around a story of a late 19th century explorer who managed to find all of these amazing undiscovered relics which he recorded in his journals and of course which mysteriously vanished. The story has been quoted several times on social media sites and this is the direct quote that you will likely see accompanying these photos.
“British explorer Alfred Isaac Midleton traveled the farthest reaches of the globe during the late 19th century in search of zoological, botanical and archeological wonders. These recently unearthed photographs help to shed light on some of his amazing discoveries during several missions to then-uncharted regions of Southeast Asia, Africa and the Amazon rainforest. Unfortunately, all of his journals and scientific writings were lost in 1901 when Midleton and his team vanished during a Sumatran expedition to uncover the fabled Lost City of Dawleetoo. Due to these tragic events, scant details are known about what you see in these fascinating photos.”
Image Source: “The Curious Journeys of A.I. Midleton”
A nightmare for future researchers.
Let us address some obvious issues first. A quick Google search will show there is no explorer from Britain by the name of Alfred Isaac Middleton. A.I. Middleton seems to be a play on words. In this case A.I. refers to artificial intelligence. Next if all of his material was lost, where did the photos come from? The script is perfectly made for an internet viral sensation and that is what it is going to become. Recently we have seen remarkable increases in complexity of artificial intelligent created imagery. There are countless apps that you can download and create artwork from. This seems innocent enough and harmless, however the user has to remember these systems are using your input to learn. Why does that matter ? It matters because the progress is exponential, what started as simple line drawings and emulations of art styles has now started to approach realism. In short what that means is soon we will not be able to trust not only still images, but also video. As researchers this is very concerning, as we come across countless examples of deliberate hoaxes created with Photoshop or other editing software there are always telltale signs that we can find. However by creating these works with artificial intelligence tools, each example gets progressively better until it will be hard or next to impossible to classify as machine or artificially created.
This work may be traced back the DALL-E2 AI system which like most AI art generation sites encourages users to use it to create art based on key words and phrases. Such as, “An astronaut riding a horse on the moon with a chunk of cheese in photorealistic style etc”. This may seem like complete trivial nonsense, but the AI is learning to interpret this and creating it as commanded. This particular system seems to be focusing on human creation and mastering photo-realism. Unlike real artists AI can cut, copy and paste as well as edit from billions of reference images at speeds only limited by their processors and in time that will be instantaneously. Soon in my opinion as well as others, will represent a massive issue for everyone who depends on social media for their news but here we are talking about cryptozoologists, UFO researchers, and other paranormal researchers.
Ancient Astronaut Theorists say "NO."
Today we are threatened by bots which have been programmed to look for articles that go against certain preprogramed criteria. Articles that have issues with Fact Check for example and that may be flagged. Comments may be disabled and posters may be banned from certain sites. It's bad enough when humans began to censor the unknown, it's worse when bots are used in a way that blurs these lines. For example, will bots recognize AI generated art as simply art? Will they flag those images as false? Or , will AI fool itself? Many may think that this is alarmist thinking however this is an issue which has been coming for a very long time. As a background to understand this it's important that we look back. in the 1950s computer scientist Alan turning created a test for machine intelligence which companies around the globe have tried to pass for years. There has been success passing this test with chat bots however those results are controversial. Recently though in 2022 there have been new developments, a new class of program called large language models. These programs are said to emulate actual human thought and not simply be repeating strings of words that may or may not make sense. The most famous of these large language models would be googles LaMDA which it’s creator claims is not only intelligent but has developed sentience.
If we look at this progress, we can see that these AI systems seem to be getting better and doing so very quickly. Right now almost anyone can harness the power of these systems to create any image that they want. Curiously enough, the image created cannot be used or sold, this makes sense to me as the image would be owned by its creator. Whether its creator is human or rather the product of a human created AI program, this to me illustrates the true function of sites like this, to use human curiosity as a way to improve machine learning. How will this affect cryptozoology? How will this affect your research and all other paranormal research? One needs to only look around at the plethora of hoaxed and faked images which circulate yearly on every social media site. Many of these are known hoaxes yet they are still posted relentlessly over and over again with a small number of us speaking out against them. Here we see something far more dangerous than a copy and a paste job.
In the case of A.I. Middleton we see a clear intent to lie, to perpetrate a known hoax using a tool, in this case an artificial intelligence tool. In this particular case not only are the images faked, there is an entire back story which has been contrived. A background is created for this fictional character much like any work of fiction, a plot is given, as a result a very confusing assortment of images all in the same style all showing the exact same artifacts for now str created. For now we can recognize the telltale signs of an AI artist but in several months, in several years, in a decade? We won't know what real is when we see it on social media. So the next time you see someone post a picture of Bigfoot, or a UFO, or ghost, you're going to have to look very hard, you're going to have to look very deep into the image. Is there something a little off? If so you may have found the fingerprints of not a Photoshop job but something far more troublesome.
The Eyes Tell the Tale
Note the eyes, so far this is one of the most easily identified artifacts of machine created art representing humanoid characters. The AI does not yet seem to understand symmetry or predictive compositions. This will likely chan