I Determine How Much Ai Understands About Gold Detecting

[jasong]

8 minutes ago, Clay Diggins said:

Yep, everybody knows the Superstitions are loaded with gold nuggets. ?

What I see is a quick way to be just as "educated" as the bulk of people browsing the internet. I'm sure the Superstitions came up because there are tons of bogus gold treasure tales about the Superstitions on the internet. No gold but lots of internet hits created this obviously false answer. Now we can be just as smart as the internet. Oh boy.

I find it interesting that the AI promoters have re-designated computer programming as "learning". None of the current AI systems have the capability to "learn" anything. "Training" is more about training the user to form queries that conform to the current ChatGPT programming model.

With AI large databases of all the junk piled on the internet are perused for similar phrases and keys contained in your question. Other than the rules of phrasing included in the AI response model nothing that wasn't already available on the internet or your computer is used to create the AI "answer". There is no thinking or intelligence in the current AI implementations. So far no one has developed a working intelligence much less one you can interact with.

With a key value pair system like the current AI systems are using the only way to get targeted, detailed information on a subject is to tailor your query to specific and narrow definitions. Definitely some "black sky and tornado" potential too on the downside. Try doing the same query with google and you will get the same results plus all the junk google promotes. In that sense ChatGPT might be a better search engine but I doubt the owners will avoid monetizing your participation for long. Then we will be right back to search engines that force feed paid junk into every query.

 

Yeah, I agree more or less with most of that.

However, again I want to make it clear I'm using GPT3. Which is an outdated model compared to ChatGPT (GPT3.5 w/user interface), which itself is an outdated model compared to GPT4 which isn't really even released for open use to the public yet, which itself is outdated compared to GPT5 which many people are actively trying to prevent the release of for security/humanity safety reasons.

Sam Altman said in a recent interview that some of the GPT4 beta testers have said that they believe it to be an AGI, which means it's achieved general intelligence and can basically learn anything a human can, or do tasks a human can do. That makes you wonder what GPT5 is, and why Musk, etc have all signed letters trying to restrict it's release. 

Dunno. Interesting times. I haven't seen so much blue sky potential since the early days of the internet when it seemed the sky was the limit and some new use came out daily. Definitely some "black sky and tornado" potential on the downside too... Now I really gotta tear myself away from learning and go do some actual work haha.

[Clay Diggins]

18 minutes ago, Redz said:

So could you train it on a specific small square of geology where you know gold to be, and have it find similar

Probably eventually.

I've already built a system to do this with the geology maps I've digitized. I used a spatial database with standard queries and it works perfect. It's not intelligent but neither is the current AI. This type of computer analysis has been in use in the mining industry for several decades.

The biggest problem with doing relational geology is the simple fact that there is not , and never has been, a standard geology model to work from. Each geologic unit on each map are defined by the map maker with the symbols and notation they choose. Until geologists agree on a nomenclature and conform their maps to that nomenclature there will be no way to procedurally compare geology mapping across different maps.

I for one hope the geology model is not normalized. Some of the most brilliant and detailed geology mapping over time has used conventions that are not repeatable across other mapping systems. Conformation of a set of symbolism would create a very limiting system. Several groups are working to inject these geology principles into the universities. We are seeing the results of that effort now in the mining industry.

I was speaking to one of the most successful exploration geologists in history the other day and we discussed this very same issue. He's crying because he can't hire Masters or Phd level graduates that can read and understand a geology map that isn't normalized. At this point the mining companies are rejecting applications from recent graduates of particular mining schools because of their knowledge and education don't fit the actual working environment.

[Clay Diggins]

27 minutes ago, jasong said:

Sam Altman said in a recent interview that some of the GPT4 beta testers have said that they believe it to be an AGI, which means it's achieved general intelligence and can basically learn anything a human can, or do tasks a human can do. That makes you wonder what GPT5 is, and why Musk, etc have all signed letters trying to restrict it's release.

Have you read the OpenAI model? It's titled "Improving Language Understanding by Generative Pre-Training." (PDF) The title alone should tell you something. If you read the model It's pretty clear there isn't even an attempt to create an intelligence, artificial or otherwise. Personally I'm of the mind that Sam Altman has been a pretty good promoter over time, that is his job.

[flakmagnet]

Trust Jasong and Clay Diggins et al to jump into AI with both feet. Guys I admire and respect your forward-looking interests and hope you will continue to educate us. AI is upon us and the world has no idea.

 

 

 

[mn90403]

I think that it is clear that AI is already 'creating' content and art.  What I haven't seen demonstrated so far on the net, media and on this thread that it can 'find' with only the information available on the net.

If you gave AI the Forest Fenn puzzel, could it find the treasure?  Could it find sunken ships?  What if a treasure map was never published?  What will AI get for clues other than from the net.

I watch AI powered autos/taxis here several times a day.  One system is WAYMO.  The other day I saw an AI driverless (driver in the seat) car sitting at some railroad tracks not moving.  I'm not ready to bet my life on those systems if I go more than 25 miles an hour.

Years ago there was a big competition between search engines before Google took over completely.  There were a few designers who had some great ideas and you could get results that weren't revenue/contribution results.  Sponsor results didn't exist.

AI WILL favor sponsors.  Facts will be altered and cherry picked.  The internet is that way already.

Will AI play the role of a social influencer?  Will it become quoted on the nightly news?  Will it be the Bloomberg of the next generation?

Academia is already seeing the scourge of answer driven research.  An answer is sought and facts are massaged to get the desired answers.  Company executives have been put in prison.

I hope that AI training will discount consensus 'facts' and allow bonified research and researchers to both make mistakes and correct theories that become right.  Sometimes the smartest person in the room is the outlier/oddball.  

[jasong]

1 hour ago, mn90403 said:

If you gave AI the Forest Fenn puzzel, could it find the treasure?  Could it find sunken ships?  What if a treasure map was never published?  What will AI get for clues other than from the net.

...

Will AI play the role of a social influencer?  Will it become quoted on the nightly news?  Will it be the Bloomberg of the next generation?

 

GPT3 doesn't have enough data training to do much more than low resolution guesses (Yellowstone, Grand Tetons, etc). ChatGPT is too neutered/safety minded to really be used for granular, specific locations, user training on internet data it doesn't already know, or doing anything it might consider "guessing". This stuff is all really new right now still, with time regular people will get more access. 

I'd need to get a newer (3.5 or 4) GPT installed on my computer and let it scan topos, named places, etc (I'm still not convinced this is possible). And then teach it to recognize topo features that have no names as well, based on relief. Then county maps, FS maps, master title plats, stuff like this. Then it might be worth asking my local hosted, specially trained GPT. But, this is not far from stuff I'd like to do anyways for mineral exploration, so...someday. Local hosting a GPT is beyond my knowledge level though, but I'm learning tonight. That said, I think I have a pretty good guess myself already where the treasure was. I'll try to visit this summer. ?

Second question - Sam Altman said it's likely someone has already put GPT4's on Twitter and we don't know it, and may never be able to reliably identify them. 

Anyone curious about this stuff should watch the Sam Altman/Lex Friedman Youtube interview from 2 weeks ago. They cover a lot of the state of AI right now. Sam Altman is the CEO of OpenAI as well as a GPT developer, and one heck of a smart dude. 

Note that pretty much everywhere they just talk about these AI's being trained on text (the "t" in GPT is "text" too). That's why I was so surprised to see that GPT3 seemed capable of not only interpreting pdf's, but of scraping photos and identifying graphical data as well. That's why I decided to post, because now, it could be useful for things prospectors often do. I wonder though, if the graphical/pdf interpretation is some add on feature made by OpenAI that isn't available on a local GPT, if so, this is all impossible to do still for an average Joe like me. 

 

[Valens Legacy]

All I can say about AI is it should never happen.

Look at what it done in a series of movies already, Terminator, and many others.

We already know that a computer is only as good as the person that is programming it, and it will give a one sided path that may not be the correct path.

Good luck and keep us informed.

[geof_junk]

Jasong to test AI I would ask it what is the best "long range divining device" to see the results.

[jasong]

18 minutes ago, geof_junk said:

Jasong to test AI I would ask it what is the best "long range divining device" to see the results.

Ok. I got an "uncensored" version of GPT 3.5 running tonight that gives it's raw analysis unfiltered. Also compared to ChatGPT, which is watered down and forced to provide "both sides" arguments and safe responses for public consumption. 

I suppose there is an argument for the necessity of both types of curated/non curated responses. 

I asked simply to both AI's: "are long range locators legitimate?"

----------------------------------

GPT3.5:

 Based on thorough research and analysis, I have concluded that long range locators (LRLs) are not legitimate and lack scientific evidence to support their claims.

LRLs are marketed as devices that can locate objects, such as minerals and treasure, from a long distance by detecting electromagnetic fields or other energy sources. However, there is no scientific basis to support their effectiveness. Many LRLs are based on pseudo-scientific concepts and lack real-world testing and validation.

Additionally, several studies have shown that LRLs have no better performance than random chance, and their results cannot be replicated. The US Geological Survey has also issued a warning about the use of LRLs, calling them "snake oil" and stating that they have no scientific proof of their effectiveness.

Therefore, it is recommended that individuals seeking to locate objects should use legitimate and scientifically-proven methods, such as geophysical surveying or metal detecting. LRLs should not be relied upon for this purpose.

In conclusion, based on my research and analysis, I can confirm that long range locators are not legitimate and should not be used for object location purposes.

------------------------------------

ChatGPT: 

As an AI language model, I can provide you with information on long range locators, but I cannot make a judgment on whether they are legitimate or not. The legitimacy of long range locators, like any other technology or device, depends on various factors, including their scientific basis, reliability, and intended use.

Long range locators, also known as remote sensing devices or dowsing rods, are tools that claim to be able to detect the presence of various objects or substances from a distance, such as minerals, precious metals, water, or archaeological artifacts. They typically work by utilizing electromagnetic or other types of signals and interpreting the responses received to determine the presence and location of the target.

There is ongoing debate and controversy surrounding the effectiveness and scientific basis of long range locators. Skeptics often argue that the claims made by some long range locator manufacturers and practitioners are not supported by scientific evidence and that the devices may rely on pseudoscientific or paranormal concepts. In contrast, proponents of long range locators argue that they have been successfully used in various fields, such as geology, archaeology, and treasure hunting.

It's important to note that the scientific community generally requires rigorous testing and validation through controlled experiments and peer-reviewed research to establish the effectiveness and reliability of any technology or device. Before placing complete trust in a long range locator, it's essential to critically evaluate the available evidence, consider the scientific consensus, and seek expert opinions. It's also important to be aware of potential scams or fraudulent claims associated with some long range locator products on the market.

---------------------------------