Last month on my news feed I came across an article that caught my eye. The article was reporting on a scientific paper submitted by a group of computer researchers from the UK and Canada on the online science portal Arxiv, which were warning about the impending probability that the massive proliferation of large language models or LLM’s —the technical term for the artificial intelligence programs which have become so popular lately —like Midjourney or ChatGPT — which are ‘trained’ through incredibly large datasets ‘crawled’ from the Internet (artists and creators of original content call that ‘stealing’ since it is largely done without any consent or compensation) will inevitably collapse onto themselves like a snake eating its own tail, once their developers eventually run out of original —and human generated— material to feed to their hungry creations.
Once that happens, those developers will have no choice but to train their LLMs on content and dataset previously generated by other LLMs, thus triggering an irreversible negative feedback loop in which the results provided by the A.I.’s become more and more derivative, as well as less reliable in their factual content; until it eventually all these algorithms can regurgitate is nothing more than digital garbage.
Specifically looking at probability distributions for text-to-text and image-to-image AI generative models, the researchers concluded that “learning from data produced by other models causes model collapse — a degenerative process whereby, over time, models forget the true underlying data distribution … this process is inevitable, even for cases with almost ideal conditions for long-term learning.”
“Over time, mistakes in generated data compound and ultimately force models that learn from generated data to misperceive reality even further,” wrote one of the paper’s leading authors, Ilia Shumailov, in an email to VentureBeat. “We were surprised to observe how quickly model collapse happens: Models can rapidly forget most of the original data from which they initially learned.”
In other words: as an AI training model is exposed to more AI-generated data, it performs worse over time, producing more errors in the responses and content it generates, and producing far less non-erroneous variety in its responses.
~The AI feedback loop: Researchers warn of ‘model collapse’ as AI trains on AI-generated content (via venturebeat.com)
The news was interesting to me on two different levels. On the one hand, as a designer and digital artist, I have a vested interest in how A.I. programs are threatening the livelihood of people like me, as more and more companies are jumping into the A.I. trend and preparing to fire their staff, because they are under the delusion that creative labor can now be easily replaced with these ‘miracle machines’ — managers, after all, have always believed all designers ‘do’ is just press a few buttons on the computer and Voilà! A product is made.
But the other reason I found the concept of model collapse so interesting, is because it can also be used as a good analogy to what’s been happening to the field of Ufology in the last few decades.
When David Grusch, the new golden child of Disclosure, irrupted on the public scene as the latest whistleblower with incendiary claims with regards to what the U.S. government has supposedly hidden from the American public (and the rest of the world) for decades —recovered alien craft (not just one but several) and alien bodies which have been secretly analyzed by private companies without any Congressional oversight, under a clandestine program which goes all the way back to the end of World War 2; secret dealings between government officials and non-human intelligences; a looming direct or indirect threat to mankind from these entities; un-checked assassinations of civilians in order to perpetuate this secret— I could not help but wonder if we’re not witnessing a chain reaction caused by a ’derivative’ and highly degraded ‘dataset’ (to borrow terms from computer science) provided by a very closed group of individuals which has effectively hijacked the UFO narrative since 2017; the effects of which has even managed to permeate the corridors of Washington D.C.
Causes for this concern began to appear almost immediately, when independent researchers and UFO enthusiasts began to point out to the obvious ties between Grusch and the most prominent voices in this new UFO scene: Eric Davis, Chris Mellon, Lue Elizondo, Jeremy Corbell, etc.
Take for instance Grusch’s statement that an alien vessel crashed in Italy in the 1930s and was retrieved by the U.S. government after the fall of Mussolini —with the help of the Vatican, no less. Is it too preposterous to assume Grusch first heard of this improbable story from Elizondo, who we know went to a UFO conference in Italy with Tom DeLonge after he resigned from his Pentagon assignment and entered TTSA, which was later represented as an ‘official’ meeting with the Italian government on the first season of their TV show Unidentified?
And with regards to the claims of super-secret projects focused on the reversed engineering of alien technology, might he not be just referencing the controversial Wilson Memo which was written by Davies and first came into light after it was found in the archives of the late Dr. Edgar Mitchell after he passed away?
When Grusch mentioned to journalist Ross Coulthart, in his interview for News Media, the notion that the recovered ‘alien artifacts’ seem to display peculiar isotope ratios, that to me seemed a telltale indication that he’d been talking with someone like Garry Nolan, who along with Jacques Vallee were the first ones to point out the notion that the purported material samples from UFOs they’ve independently analyzed seemed to be neither meteoric in origin (i.e. from outer space) nor naturally occurring in our planet, thus leading them to suspect someone or something had ‘re-engineered’ these metals at the molecular level —for reasons nobody has seemed to deduce, as far as we know.
You see, within the model collapse analogy it doesn’t matter how smart, experienced or well-connected a person is. The tainted dataset will eventually ensure an erroneous outcome in the long run. Garbage In = Garbage Out
My point being is this: the reason all these dots seem to be connecting so effortlessly all of a sudden, might be because we’ve been hearing the same story coming from the same people, without any way to verify their evidence: Rumors and gossip taken as incontrovertible facts and disseminated at light speed by an army of sock puppet accounts and sycophants on social media —who take great joy in having been ‘chosen’ for the task of paving the road to the golden promise of Disclosure— and the rumors get eventually picked up by websites and news organizations who care more about clicks and ad revenue than fact-checking nowadays. Thus, entering an irreversible situation in which there is no practical way to gauge if the ‘trained dataset’ is reliable or not; and the few who dare to speak up and demand tangible evidence are singled out as ‘traitors’ or debunkers, even if one does not share the opinions of ‘pure-blood’ skeptics like Mick West, who firmly believe all UFO reports could be easily explained in simple terms with enough information.
Grusch, in this scenario, would not be then perceived so much as a whistleblower rather than an errand boy with credentials impeccable enough to catch the attention of the public and Washington. And who knows? He could inadvertently also be the one deliberately sent to cause such an embarrassing blowup, so that all who might have had a modicum of interest in the UFO subject after the New York Times article of 2017 was published decide henceforth to back away in disgust and never again touch the topic with a ten foot pole; leaving the field with only the true believers, the skeptics who get off by making fun of them, and the spooks who exploit the mystery for their own particular goals.
And if you think I’m writing nonsense, and Grusch did manage to gather genuine evidence outside the small group who’s been shaping the UFO narrative in the XXIst century, my reply is we can still find other signs of the ‘model collapse’ I am worried about in the tweets of the same lawmakers organizing the big bru-ha-ha taking place tomorrow in D.C. have been sharing prior to the event. Just take a look at this Tweet by Representative Anna Paulina Luna (R-Fla) in which a cursory glance by anyone who’s have paid enough attention to what’s been going on in the last few years should cause them to cringe.
Aside from some ‘classic’ UFO photos and the grainy still images of the FLIR and Gimbal videos we’ve seen ad-nauseam since 2017, one can also glimpse a copy of the infamous ‘Batman balloon’ photo which was first shared by The Debrief after it was leaked to them by an ‘anonymous source’, and was later thoroughly debunked —first by online researchers and then by the government UFO Task Force— as well as an image taken from the ‘triangular UFO’ video leaked by Jeremy Corbell which was first explained by Mick West as lens artifacts (bokeh) —and later confirmed by AARO, the current official UFO research group in the US government.
Just who is instructing Luna and Tim Burchett (R-Tn) —another lawmaker who has found in the UFO topic a great promoting platform, and who seems to believe reptilian aliens control the world— about this topic?
Or is their evident ignorance about the nuances in the topic the result of the toxic ‘training model’ everyone has been subjected to?
This could very well be the article that ages the worst in my career as a UFO commentator. Perhaps tomorrow we will all be shocked to discover we are all living in a brave new world in which the existence of non-human intelligences who regularly interact with us on our own planet is largely accepted; in which case I will gladly congratulate Grusch and all those who pulled the strings to make it happen —even if I still question their cloak-and-dagger methods.
But if all we get from the House Oversight Committee hearing tomorrow is yet another rehash of David Fravor and Ryan Graves’s testimony —something we’ve heard a hundred times before on a hundred different interviews for the last five years— and a few new sensational claims from Grusch with nothing to back them up outside of his word, then I’m afraid the model collapse of this current UFO narrative will be unstoppable, and we will be forced —with good reason— to start all over… again.
…I only hope that next time we make sure not to trust the bots so much.
