Geolocating with GPT4

Mick West

Administrator
Staff member
Similar case reported to MUFON:
20170224-093603-v4fva.jpg
MUFON case 82114

2017-02-16
2017-02-15 12:06PM
Dark, 4 figures standing on top of clouds, all deserted then these figures standing in the middle of the cloud
Lodz, , DE

Flying from Warsaw to London Heathrow, looking out the window to my left.
Out of the corner of my eyes I saw these dark, thin figures in the middle of a cloud.
It definitely wasn't a building because we were too high up so I thought they were mini tornadoes
There were 4 figures: 2 tall ones and 2 slightly small ones. They were very close together and they weren't moving at all. They were dark, had a thicker upper body and a thinner lower body. One of them looks like they had wings.
I was in awe and shocked at how it was all deserted then there were these random weird objects just standing on top of a cloud. I took photos and showed my parents and brother the photo and they couldn't believe their eyes. I was also a little bit excited because I have never seen anything like this before.
Flew past the object(s) then it disappeared.
Content from External Source
If you review the preceding posts it's quite obvious this is a just another set of power station cooling tower plumes. The enterprising reader should even be able to track down which one.

I was just tinkering with this, and thought I'd ask ChatGPT4 to help. My first questions was kind of simple:

which power stations have six or more cooling towers


Power stations with six or more cooling towers are typically large, coal or nuclear power plants, as these types of facilities require significant cooling capacity. Here are some examples of power stations with six or more cooling towers:
  1. Drax Power Station, UK: Drax is a large coal-fired power station with a total of twelve cooling towers, making it one of the largest in the United Kingdom.
  2. Ratcliffe-on-Soar Power Station, UK: Ratcliffe-on-Soar is another coal-fired power station in the UK, which has eight cooling towers.
  3. Cottam Power Station, UK: Cottam is a coal-fired power station in the UK, which has eight cooling towers.
  4. Ferrybridge Power Station, UK: Ferrybridge is a coal-fired power station in the UK, which has eight cooling towers.
  5. Eggborough Power Station, UK: Eggborough is a coal-fired power station in the UK, which has eight cooling towers.
  6. Cattenom Nuclear Power Plant, France: The Cattenom Nuclear Power Plant in France has eight cooling towers, making it one of the largest nuclear power plants in the world in terms of cooling capacity.
  7. Dukovany Nuclear Power Station, Czech Republic: The Dukovany Nuclear Power Station has six cooling towers.
Content from External Source
Then I noticed the "Lodz" location, so given that GPT4 seemed to know about power stations, I got more specific

name some big power stations near Lodz, poland


Lodz is a city in central Poland, and there are several notable power stations in the vicinity. Here are some of the major power stations near Lodz:
  1. Bełchatów Power Station: Located about 60 kilometers (37 miles) southwest of Lodz, the Bełchatów Power Station is the largest coal-fired power station in Europe and one of the largest in the world. It has an installed capacity of approximately 5,300 MW.
  2. Pątnów-Adamów-Konin (PAK) Power Complex: The PAK Power Complex is located about 130 kilometers (80 miles) west of Lodz and consists of three power stations: Pątnów, Adamów, and Konin. The combined installed capacity of these power stations is around 2,000 MW, with the majority of the electricity generated from lignite.
  3. Kozienice Power Station: Situated approximately 180 kilometers (111 miles) east of Lodz, the Kozienice Power Station is one of the largest coal-fired power stations in Poland. It has an installed capacity of around 4,000 MW.
  4. Opole Power Station: Located about 200 kilometers (124 miles) southwest of Lodz, the Opole Power Station is a coal-fired power plant with an installed capacity of over 3,500 MW.
  5. Łagisza Power Station: Approximately 140 kilometers (87 miles) south of
Content from External Source
(GPT4 is currently limiting responses, but you can type "continue" to get the rest)

I realized I was doing work that a computer could do, but previously had been unable to do so. How far could I push it?


name some big power stations visible from the plane when flying from Warsaw, Poland, to London, UK



When flying from Warsaw, Poland, to London, UK, the flight path typically passes over several countries, including Poland, Germany, the Netherlands, and the UK. Along this route, you may be able to spot several large power stations, depending on the exact flight path and visibility conditions. Here are a few notable power stations that may be visible during the flight:
  1. Bełchatów Power Station (Poland): As mentioned earlier, this is the largest coal-fired power station in Europe and could be visible shortly after departing from Warsaw.
  2. Turow Power Station (Poland): Located near the border between Poland, Germany, and the Czech Republic, the Turow Power Station is a large lignite-fired power plant with a capacity of around 1,494 MW.
  3. Jänschwalde Power Station (Germany): Situated close to the German-Polish border, Jänschwalde is a coal-fired power station with an installed capacity of approximately 3,000 MW.
  4. Neurath Power Station (Germany): Located near the German city of Grevenbroich, the Neurath Power Station is a lignite-fired plant with a capacity of over 4,000 MW.
  5. Maasvlakte Power Stations (Netherlands): Situated near Rotterdam, the Maasvlakte power stations include several coal and gas
Content from External Source
Nice! I started to look these up in Google Earth, dropping pins on them, then again realized I was doing grunt work. So.....


Make a KML file that includes the locations of those

Creating a KML file requires a text editor to write the KML code, which is an XML-based format. You can create a simple KML file with the mentioned power stations by following the example below. Copy the content and save it as a .kml file (e.g., power_stations.kml). Then, you can open the file using Google Earth or any other compatible software to visualize the locations.
Content from External Source
Code:
<?xml version="1.0" encoding="UTF-8"?>
<kml xmlns="http://www.opengis.net/kml/2.2">
  <Document>
    <name>Power Stations</name>
    <Placemark>
      <name>Bełchatów Power Station</name>
      <Point>
        <coordinates>19.320481,51.261267,0</coordinates>
      </Point>
    </Placemark>
    <Placemark>
      <name>Turow Power Station</name>
      <Point>
        <coordinates>14.956332,51.165126,0</coordinates>
      </Point>
    </Placemark>
    <Placemark>
      <name>Jänschwalde Power Station</name>
      <Point>
        <coordinates>14.528196,51.862938,0</coordinates>
      </Point>
    </Placemark>
    <Placemark>
      <name>Neurath Power Station</name>
      <Point>
        <coordinates>6.558869,51.035207,0</coordinates>
</Point>
</Placemark>
<Placemark>
<name>Maasvlakte Power Stations</name>
<Point>
<coordinates>4.027431,51.973903,0</coordinates>
</Point>
</Placemark>
</Document>
</kml>

Loaded that into GE, and added a line:

2023-03-30_04-32-57.jpg

Impressive! That saves a lot of work, and suggests that using AI for geolocation is going to be a very useful tool.

And in this instance, Janschwalde looks like a good candidate
 
With another case:
https://www.metabunk.org/threads/de...t-was-giving-off-bright-lights-and-orbs.7021/



Initially just asked it with no real help:


What is this: An airline passenger claims to have captured images of a UFO giving off 'bright lights and orbs' on the ground near the top-secret Area 51 military base. The witness was on an American Airlines flight from San Jose, California to Houston, Texas, when he took a series of pictures of a giant metallic disc in the Nevada desert. He said the plane was flying near Luning and Gabbs on October 30 when he noticed 'blinding bright light' in the desert below.


The passage you provided describes a claim made by an airline passenger who alleges to have captured images of a UFO (Unidentified Flying Object) emitting bright lights and orbs near Area 51, a top-secret military base in Nevada. The passenger was on an American Airlines flight from San Jose, California, to Houston, Texas, when they took pictures of what they believed to be a large metallic disc in the Nevada desert. They reported noticing a blinding bright light in the desert below while flying near Luning and Gabbs on October 30.
While such claims often generate interest and speculation, it is important to approach them with skepticism. There is no definitive evidence to confirm the existence of extraterrestrial life or UFOs associated with aliens. Many sightings can be attributed to misidentifications of natural phenomena, conventional aircraft, weather balloons, or other explainable occurrences.
Content from External Source
Find a solar power facility that might explain it

The passage describes a sighting of bright lights in the Nevada desert. One possible explanation for this could be a solar power facility reflecting sunlight. Nevada has several solar power facilities, and one that could potentially be visible from the flight path between San Jose, California, and Houston, Texas, is the Crescent Dunes Solar Energy Project.

The Crescent Dunes Solar Energy Project is located near Tonopah, Nevada, approximately halfway between Las Vegas and Reno. This facility utilizes concentrated solar power (CSP) technology, which uses thousands of mirrors to focus sunlight onto a central tower containing molten salt. The intense heat generated by the focused sunlight is used to create steam, which powers turbines to produce electricity.

The mirrors, or heliostats, in the Crescent Dunes facility might cause a bright reflection that could be visible from an airplane, especially if the sunlight is focused in a particular direction. This explanation is speculative, but it's possible that the bright light observed by the airline passenger was a reflection from a solar power facility like Crescent Dunes, rather than a UFO.
Content from External Source
Bingo! Just needed a bit of help.
 
the AI isn't actually smart. It provides a natural-language interface to a huge trove of data.
Yeah I use it at work as a sort of assistant, it's great at sort of research and presentation.

Micks example is a great demonstration as to why a human mind has to do the difficult bit.
 
Yeah I use it at work as a sort of assistant, it's great at sort of research and presentation.

Micks example is a great demonstration as to why a human mind has to do the difficult bit.

Yes, but the AI's next answers to the same initial question by Mick may well connect the UFO sighting to solar power plants, or even to a particular solar power plant.

Maybe someone could test it?

What the AI can't do is unprogrammed, unemulated and/or non-stochastic generation of meaningful text.

In other words, it cannot consistently generate new meaningful text or consistently answer meaningfully to new questions on existing texts.

Since humans for the most part don't choose to analyze existing data probingly nor create new meaningful text content although capable of doing both (and which is a pity and a generational educational challenge for homes and schools), GPT4 and other large language model AIs are really good at simulating commonplace (read: dumb) human text-generation. And only getting better and doing it faster than us.

Which is impressive in its own right.

GPT4 is a Google on nitro.
 
Last edited:
So when AI came out with a new and better way to multiply matrices together
https://www.nature.com/articles/s41586-022-05172-4
Improving the efficiency of algorithms for fundamental computations can have a widespread impact, as it can affect the overall speed of a large amount of computations. Matrix multiplication is one such primitive task, occurring in many systems—from neural networks to scientific computing routines. The automatic discovery of algorithms using machine learning offers the prospect of reaching beyond human intuition and outperforming the current best human-designed algorithms.
What text book were they getting their answer from?
Answer = none, it came to the answer itself

The truth is we (humans) have no idea exactly how AI comes to its conclusions, we make simplified assumptions on how its working (involving said matrices ;) ), but its too complicated to really know. If you think we know how it works, now explain how your human conscious mind sitting behind your keyboard came to that conclusion? (neurons fired passing electricity blah blah), or exactly what the weather will be here in a months time at 10am

Disclaimer: I'm not claiming AI is sentient or intelligent.
Though obviously they are intelligent, even a cockroach is intelligent
AI's now easily pass the turing test, we need a better test, the future is going to become extremely murky
 
So when AI came out with a new and better way to multiply matrices together
https://www.nature.com/articles/s41586-022-05172-4

What text book were they getting their answer from?
Answer = none, it came to the answer itself

The truth is we (humans) have no idea exactly how AI comes to its conclusions, we make simplified assumptions on how its working (involving said matrices ;) ), but its too complicated to really know. If you think we know how it works, now explain how your human conscious mind sitting behind your keyboard came to that conclusion? (neurons fired passing electricity blah blah), or exactly what the weather will be here in a months time at 10am

Disclaimer: I'm not claiming AI is sentient or intelligent.
Though obviously they are intelligent, even a cockroach is intelligent
AI's now easily pass the turing test, we need a better test, the future is going to become extremely murky

The exact neural pathways and processes occurring in the organism 'mouse' everytime it runs away from a cat are also unknown but not the overall mechanism. No real mystery there.

Here's a good read:

204C5456-FCEB-4B30-B493-AC7769387FDC.jpeg
 
Last edited by a moderator:
What text book were they getting their answer from?
Answer = none, it came to the answer itself
not really true, that AI algorithm was programmed to search a predefined solution space
basically it had the task to find needles in a haystack and excelled at it, but there's nothing qualitatively new about the solutions it finds
 
Yes, but the AI's next answers to the same initial question by Mick may well connect the UFO sighting to solar power plants, or even to a particular solar power plant.

Maybe someone could test it?

What the AI can't do is unprogrammed, unemulated and/or non-stochastic generation of meaningful text.

In other words, it cannot consistently generate new meaningful text or consistently answer meaningfully to new questions on existing texts.

Since humans for the most part don't choose to analyze existing data probingly nor create new meaningful text content although capable of doing both (and which is a pity and a generational educational challenge for homes and schools), GPT4 and other large language model AIs are really good at simulating commonplace (read: dumb) human text-generation. And only getting better and doing it faster than us.

Which is impressive in its own right.

GPT4 is a Google on nitro.

1680431603811.png
 
Since humans for the most part don't choose to analyze existing data probingly nor create new meaningful text content although capable of doing both (and which is a pity and a generational educational challenge for homes and schools), GPT4 and other large language model AIs are really good at simulating commonplace (read: dumb) human text-generation. And only getting better and doing it faster than us.

I view it similar to the wisdom of crowds. It does have a stochastically tumbling orbit around the truth, but not through planning or design, just from sheer number of inputs, some of which will align.

I became made aware recently of a question asked of ChatGPT3 that was tangentially related to me, and basically almost every single thing that it responded with was false (it got my name correct, that's about all). Some was hilariously inaccurate, and could have been disproven by a moderately bright 9 year old counting on his fingers.
 
As another example, on another forum a user was trying to identify the film that a specific scene he recalled was from. When a bunch of movie nerds failed to come up with anything, he tried ChatGPT. Over several sessions it offered several movies, all real. It gave precise timestamps of the scene and the actors involved, all false. The user never did figure out the movie.

A good way to look at it: if something is reliably recorded in the training data, ChatGPT can usually find it much easier than most humans can navigate the more arcane aspects of search engines. However if the answer isn't reliably recorded, ChatGPT will lie to you blatantly and boldly.
 
Last edited:
A good way to look at it: if something is reliably recorded in the training data, ChatGPT can usually find it much easier than most humans can navigate the more arcane aspects of search engines. However if the answer isn't reliably recorded, ChatGPT will lie to you blatantly and boldly.
I heard a lecture on "expert systems" over 30 years ago, then all the rage in AI. The main difference between those and humans is that they don't have any clue when they've crossed the boundary of their expertise and their performance drops sharply. That's precisely the behaviour you describe ChatGPT having. (I heard of another anecdote where it completely fabricated the description of a small (real) museum.)

You really need to think of its output as "the Internet says..." and cross-check it.
 
I heard a lecture on "expert systems" over 30 years ago, then all the rage in AI. The main difference between those and humans is that they don't have any clue when they've crossed the boundary of their expertise and their performance drops sharply. That's precisely the behaviour you describe ChatGPT having. (I heard of another anecdote where it completely fabricated the description of a small (real) museum.)

You really need to think of its output as "the Internet says..." and cross-check it.

An apt encapsulation.

I also like the term 'stochastic parrot'.

Article:
"A system for haphazardly stitching together sequences of linguistic forms it has observed in its vast training data, according to probabilistic information about how they combine, but without any reference to meaning: a stochastic parrot.”
 
ChatGPT will lie to you blatantly and boldly.
The technical term for that is "hallucination".
Article:
In artificial intelligence (AI), a hallucination or artificial hallucination (also occasionally called delusion[1]) is a confident response by an AI that does not seem to be justified by its training data. [..]

Such phenomena are termed "hallucinations", in analogy with the phenomenon of hallucination in human psychology. Note that while a human hallucination is a percept by a human that cannot sensibly be associated with the portion of the external world that the human is currently directly observing with sense organs, an AI hallucination is instead a confident response by an AI that cannot be grounded in any of its training data.[2] AI hallucination gained prominence around 2022 alongside the rollout of certain large language models (LLMs) such as ChatGPT.[4] Users complained that such bots often seemed to "sociopathically" and pointlessly embed plausible-sounding random falsehoods within its generated content.[5] Another example of hallucination in artificial intelligence is when the AI or chatbot forget that they are one and claim to be human.[6]

By 2023, analysts considered frequent hallucination to be a major problem in LLM technology.[7]
 
Back
Top