Leaderboard

My name is Shelley and I recently had an encounter on a BFRO expedition. I have always been interested in this subject, but this experience has motivated me to learn as much as I can about this subject. I appreciate being added to the Bigfoot Forum and I hope to learn from some of the experienced researchers here. Cheers!

Computer guy here. High scale, petabytes scale, and higher. Tens of thousands of servers and hundreds of thousands of "containers" scale. And of course loads of old and recent machine learning and AI experience. It's a tool. It's fed poorly, limited information from limited sources, sources that are biased towards certain answers and limited in many ways. I am curious as to what it would regurgitate if it were trained in all US government archives, like one had access to a trove of secret documents. For example, I use an SDK or software agent to deliver log files and metrics to time series databases, that SDK and agent have libraries written for it. When asked how to plumb up certain technologies with the SDK, some AIs will make up answers out of whole cloth, faking a reply. For me it's not a huge deal because the libraries are protected by many layers but for other tools a bad actor could write that fake library and use it for compromising systems. Unreliable. But teach it to only reply with proven libraries with digital signatures signed by some author and your answers improve. I suspect that if we ever let it in to truly hidden archives, the secrets it would out would far sooner start WW3 than expose a cover-up of cryptids, but that's the only way something like this would be revealed by AI. We would have to plumb the archives of a dozen agencies to get access to the mundane reports from the field and even then have to piece it together ourselves. Archives that won't be made public anytime soon. The current language models are trained on mundane stuff, unfortunately. But, someday, you'll stumble into some university archives only accessible via secret username and password, and pass that login information and URL to an LLM and feed and prompt it, here's a URL and the login ID is Joe, password is password123, digest all you can find and summarize the reports mentioning Bigfoot, Sasquatch and any other similar species... And only then will it start to get fun. The question is who lets one in first and who writes the first prompt to expose it.

I'm a believer in Sasquatch, and i hope to someday become a knower but I asked chatgpt if Bigfoot is real and this is what it said: is bigfoot real? ChatGPT said: No, Bigfoot is not real. There is zero scientifically credible evidence to support the existence of such a creature. Every supposed sighting, footprint, or video has either been debunked, lacks verifiability, or is anecdotal. In contrast, extensive wildlife studies, camera traps, and modern tracking technologies have failed to produce a single verifiable specimen or consistent biological trace. The idea persists because it's profitable (TV shows, merchandise, tourism) and taps into human fascination with the unknown. But from a practical standpoint, a large, elusive primate population surviving undetected in North America is implausible. It's fiction, not fact.

This caught my attention, since my topic post w/ graph showed an observable downward trend. Looking back on my data, code, and graph, I discovered I had mistook the date field as the submission date, when in fact what the Kaggle author called 'timestamp' is actually the reported sighting date. I should have caught this. The submission date is not available in the dataset I had used. The trends that AI pulled from Reddit are based on what the Redditor called an updated dataset relative to the one I used. This updated version has a submission date and a messy sighting year field (e.g., 2022, 2014-ish, 2001-2002, 1987 and 1994, 2011, etc.). The updated version also cuts off at 2021. There are other differences between the datasets, but here's what I found in terms of AI's response: Yes, there was a spike in 2012, though these were largely Class B sightings. My guess is that this comes from heightened awareness from Finding Bigfoot, which premiered in 2011. The downward trend resumed its course after the spike in 2012. Yes, there was an upward trend but it reversed around 2005. Here's my updated graph with correct labeling (LEFT) and a graph I created from the 'updated' data linked by the Redditor (RIGHT). Note that i had fewer records to graph (on the left), as I removed any records missing a date/year value (due to the witness unable to recall the encounter date). The graph on the right, since it's using the actual submission date (rather than the encounter date), had far fewer missing values (roughly 1000 more records to graph). BFRO launched in mid-late 1990s, and this is reflected in the near-zero submissions prior to then (righthand graph).