Georeactor BlogRSS Feed
Froot of the Algorithm
People obsessed with the Mandela Effect are into AI hype now. Let's explore what that means.
In 1962, Philip K. Dick wrote The Man in the High Castle, set in a parallel universe where the Axis powers conquered America. In later speeches, including If You Find This World Bad, You Should See Some of the Others, he appeared to believe that this was not a fictional world, but events in another reality:
in fact plural realities did exist superimposed onto one another like so many film transparencies
In recent years, an unusual number of people have 'remembered' that Nelson Mandela died before being released from prison, and had a televised funeral. This is not a new creation - sleuths found a mention on Art Bell's Coast to Coast AM in 2001, a reality TV show in 2003. The term "Mandela Effect" was coined by paranormal author Fiona Broome in 2009. A subreddit formed when Mandela factually died in December 2013, and it has since become a central nexus with many shared memories and discrepancies.
Here's their definition:
The Mandela Effect is when a large group of people share a common memory of something that differs from what is generally accepted to be fact.
I have to admit that some of the most common ones (the Fruit of the Loom logo, a scene from the Bond film Moonraker) are solid for me and fascinating. By 2015, mainstream internet sources (AV Club) were publishing articles like How you spell "The Berenstain Bears" could be proof of parallel universes.
The prevailing community explanation is that these memories are true to you and your original universe, and we now reside in a parallel universe. Theories fly about if this is natural, a byproduct of CERN, or something sinister and purposeful. Our current universe is often criticized as harsh (blinding white instead of yellow sun) and strangely lacking (punchlines dropped from movies). Others believe they continue to blip through universes, with facts changing again and again. So it can be distressing.
On Mastodon I follow someone who does a traveling planetarium show in Oklahoma, and kids often ask her about flat earth and this yellow sun stuff. So I don't think it's a tiny blip.
The Quest for Proof
Many Mandela Effects concern scenes from movies. Film does include deleted scenes, alternate endings, director's cuts, and alternate releases. It's possible that you've seen the Terminator 2 ending where Sarah Connor is a grandma. Theatres showed two different openings for 2018 flop Holmes & Watson. Disney+ infamously uploads movies with previously unknown edits (I read somewhere this is due to them accepting any digital copy available?). When you find a discrepancy, you can Google and confirm it with strangers. And if that clip that you 'remember' never existed, your search may turn up Mandela Effect posts.
Unconvinced, internet users began independent research. They dig up parodies from The Simpsons, reels of old commercials, interviews with actors, old usenet threads. Moments in history (like Mandela) are cross-referenced with other events from the 90s, rumors, conspiracies, etc.
The skeptical researcher believes that the Mandela Effect can be explained by any similar scene which could update your original memory. If you saw Moonraker, then a Finnish commercial, for example. But I didn't see a Finnish commercial, I remember seeing the movie on a bus during a school trip.
Even if confronted with a scene, true believers are unmoved, because they prove that we agree on a 'correct' version, or it is 'residue' that the forces of change failed to erase.
The Quest for Not Being Asked About Parallel Universes All The Damn Time
Sinbad is the one person to harness this attention to the fullest. People were asking him about a genie movie for years. In 2016 he shared photos of a TV appearance where he wore a costume, and pointed out that anyone who remembers the movie was a kid at the time. CollegeHumor filmed a reference-filled parody with him in 2017, and a radio station got him to "confess" to the movie's existence and erasure in 2018. The commenters on these videos still insist that they know the truth.
Sinbad suffered a stroke in 2020; hopefully he continues to improve.
Alt-Memories are For Kids, By Kids
Let's consider Sinbad's observation that the Mandela Effect is associated with childhood memories. It's also common for it to be distant from the person or popular discourse - you won't find South Africans absorbed into this belief that Mandela was never their president.
When Bob Dole died in 2021, yeah I was surprised that he was still alive, but no one's described an alternate 1996 where he lost the nomination or won the presidency. These events are part of sequences and cannot be widely mis-remembered - this may contribute to why there are so many rumors about celebrity deaths or disappearance from the public eye.
I am missing some Mandela effects common to 90s kids, but along lines which match up with encountering these things a little later (I saw Berenstain Bears at a friend's house and noticed the spelling; my mom notes that the series is too stereotypical in gender roles).
For the Fruit of the Loom logo, Berenstain Bears, and Froot Loops - all brands encountered by kids - I wonder if our minds have a junk drawer vector database where similar items glom together. Kids are still learning to identify words and symbols. Possibly it takes time for brains to perfect "remember a conventional spelling plus a change". As an adult you notice new things in movies and song lyrics from when you were a kid. In the 90s people started agreeing on Baader-Meinhoff to describe what happens when we start noticing a new word everywhere. So maybe the Mandela Effect is an as-yet undefined category of stuff that kids observe and remember wrongly.
In some ways the birth of the subreddit was a reunion for millennials to share weird memories from media of the 80s and 90s. Recently it's losing steam, with a bunch of kids asking if this episode of Spongebob existed in your home universe.
Can AI and TikTok cross universes?
This March, the Midjourney subreddit discussed their images of a 2001 Great Cascadia earthquake. Because it was an AI subreddit, I'd consider this LARP-ing rather than misinformation or Mandela Effect stuff.
On Twitter, among angry comments about Portland riots and stocking up on hardcopy history books, a few weirder, hastily-written replies caught my eye:
It's like a whole other world exists inside Modjouney.
It's the mandala effect on steroids… The mandala effect was created because they knew what was comming
Not to bring my flawed memory back into this, but I'm sure that I saw enough similar comments that I went investigating.
When you search Twitter, the Mandela Effect community is doing their best to test the AI most accessible to them, ChatGPT:
Decided to ask ChatGPT about a Mandela Effect topic without directly mentioning the Mandela Effect. The results were…odd. Is it pulling info from two different timelines?
Any thoughts on how ChatGPT could get this so wrong?….does it access the Akashic records & therefore the data is greater […]
This is a new one for me - apparently Akashic Records are believed to be how psychics can access facts of everything past, present, and future.
On a recent "International Mandela Effect Conference" stream, they tested ChatGPT live; all of the presenters seem familiar with this method and interested in the results: https://www.youtube.com/watch?v=3sPneby_y-c
TikTok has several subcultures around 'glitch in the Matrix' / simulation, moving between timelines, manifesting, or shift-tok. Considering how deep these go, Mandela Effect experiences are just a given fact of life. Kind of tame, really.
I found some public TikToks which use ChatGPT to discuss the Mandela Effect:
This TikTok on Amelia Earhart doesn't use ChatGPT, but explains a thought process that TikTok allows people to share information from their timelines, and Google searches only within your own timeline. I wonder if DuckDuckGo worked for her, just by not correcting her spelling?
ChatGPT and the Subreddit
Conversations with AI, confirming the alternate realities and facts, have boomed on the Mandela Effect subreddit. A recent post asks Can automod plz delete ChatGPT posts?
One defender replies:
Chatgpt repeated a memory of mine that I couldn't find anywhere using Google or on any ME boards. I realize it doesn't prove anything, but nice to see there must have been information somewhere that supported my ME.
This is interesting because ChatGPT either confirms that a lot of people agree with your false memory, or by being agreeable, helps you reconstruct a fuller memory.
Other threads describe ChatGPT as a fundamental tool, where banning it would be akin to banning internet research altogether.
No one thinks it's God. It's a good way to find out if residue exists. It is kind of like the Wayback machine since it hasn't been updated since Sept of '21.
This hints at uncertainty about how AI systems fit into the Mandela Effect's supposed replacement and residue. If the Wayback Machine is incorruptible why not use that? If it's corrupted, what makes the AI different - that it's a mind?
QM multiworlds with no assumptions of a preferred basis.
[…] the more the computer matches ME claims… The more you have to wonder.
Despite denials, it does look like a significant number of Mandela Effect believers on different platforms do want to treat ChatGPT as drawing from direct truth and not falling from a particular reality or worldview. Not only is this wrong within our own timeline, considering the flaws of an internet corpus and manipulation of RLHF, but of course there's no basis to think that the AI tapped into a deeper metaphysics during training, or in order to answer questions.
Takeaways for AI Research
- The human experience includes things which are still poorly defined or explained. Even the stuff we can explain, we're not going to get totally right.
- There are deep corpora of shared misconceptions on the internet. I think this relates to larger models' issues on the TruthfulQA benchmark.
- It's unpredictable what people with no mental model of AI will believe an AI can do - not just "I thought it could grade my students' essays!" but "I want to interrogate suspects and identify Jack the Ripper" or "explain why current event is happening, without media biases and human emotions".
- If people hold beliefs around metaphysics or psychology or truth-seeking, they will readily attribute AI responses and errors to these beliefs.
- If people think their belief is suppressed, they will seek assurances from the AI that it can agree with and operate within this belief system.
- It may be easier to begin detecting and addressing false beliefs and events in this conspiratorial, Mandela Effect sense rather than engaging in minutia of a political 'fake news' / 'misinformation' war.