Most are unlikely to emit toxic or greenhouse gasses but they're nevertheless still dangerous because they're often very deep vertical shafts that a person could stumble across and fall in. These old mines were likely closed over when they were abandoned but often their closures/seals were made of wood that has probably rotted away over the past century or so.
It stands to reason that AI would be just as effective in this situation.
That's a problem in Germany as well [1] - particularly in NRW, where most of Germany's mining activity is concentrated for centuries. About two or three times a week an old shaft collapses somewhere in Germany, leading to sinkholes - there's tens of thousands old mine shafts in the country and information on a lot of the legal ones got lost in one of the two world wars, and on top of these come quite the lot of illegal operations. Usually the damage is in some remote area, some forest or whatnot, but in some rare cases, entire buildings vanish or have to be condemned.
[1] https://www.stern.de/gesellschaft/bergbauschaeden--zehntause...
Here are some pictures https://www.google.com/search?sca_esv=f23c15b34ae4d7a1&q=lav...
A better approach would be to tune an "AI" or rather an adaptive Kalman filter variation to highlight probable shafts from airborne EM and|or ground ERT surveys:
* https://www.earthdoc.org/content/papers/10.3997/2214-4609.20...
* https://www.sciencedirect.com/science/article/pii/S001379522...
* https://nextinvestors.com/learn-to-invest/mining/electromagn...
* https://en.wikipedia.org/wiki/Electrical_resistivity_tomogra...
* Variations on this type of thing: https://www.sciencedirect.com/science/article/abs/pii/S00983...
You could cobble together a quantity estimation doing some kind of batch data association with noisy "presence" measurements, but you're probably not much better off than k-means at that point and any KF-based measurement will basically just say "Yes N>=1" because the probability is nonzero.
In a similar handwavy fashion, after picking instrument paramers and distances that might twitch on shaft responses, filters can be zeroed on regions with no shafts to see if an enhanced response can be amplified when processing the return over ground with a shaft to the surface.
Other tells for lost 'hidden' shafts might include Lidar profiles of spoil heaps .. these are clear in some cases, eroded and softened in many others.
A good many old shafts are visible from the air in any case; more so at some times of days than others - if the will is there to map them then a first pass combo of visual processing and lidar returns to map out open shafts is a good start.
OK - fits what I'd use them for
> filters can be zeroed on regions with no shafts to see if an enhanced response can be amplified when processing the return over ground with a shaft to the surface
OK - I see, look for outliers as positives, though the input signals here are not clear I can see the path. Your KF filters produce a "likely signal" and that is used downstream to actually do estimation. Probably, if I've learned anything, by plotting those "likely signals" on a map and dispatching a team with cameras.
> Other tells for lost 'hidden' shafts might include Lidar profiles of spoil heaps
Here we're outside what I'd call just "filtering". It's more like a big data science problem to model and label mounds.
I was just confused by zeroing in on "KF" in the comment above. If we're talking a big data pipeline, then yeah, KF has its place in all that.
In building exploration targets we'd pull together gravity, radiometrics, magnetics, EM, ERT, past mining records, regional surface geochemistry et al.
If tasked with finding old workings and shafts many things spring to mind .. old road and track patterns, strong vegetation zones perhaps indicating catchment from workings and dead vegetation zones perhaps due to leached processing chemicals (cyanide etc.)
But, yes, returning to the filter options there's signal and noise; when looking for the uncommon it's useful to use (say) SVD techniques to solidly lock onto common broad area signal patterns and then remove the commonplace and look for pattern in the residual noise.
https://georesglobe.information.qld.gov.au/
Go for your life ;)
The only reason they make so much money is because they disregard externalities
I think you will find many oil and mining companies have gone bankrupt(as well as software companies) and that the problem is a bit more nuanced than just treating the entire industry as a single monolithic entity.
I don't think you understand the scale of the problem... Plugging leaking wells in the US alone is estimated to cost $280b, that's 10 years of net profit for Shell or ExxonMobil. Even taking the top 5 world oil companies _profit_ from 2022 (most profitable year so far) you're short $80b, for the US alone, and that's only onshore wells.
> Researchers estimate that there are between 2-3 million abandoned oil and gas wells in the United States, and more than 117,000 of those, across 27 states, are “orphaned”—that is, uncapped, unproductive, and with no responsible party identified to manage leakage or pollution risks
https://www.cnbc.com/2022/02/24/plugging-methane-leaking-oil...
https://www.sciline.org/environment-energy/abandoned-oil-gas...
Old mines can host gravitational energy storage.
From https://news.ycombinator.com/item?id=35778721 :
> FWIU we already have enough abandoned mines in the world to do all of our energy storage needs?
"Gravity batteries: Abandoned mines could store enough energy to power ‘the entire earth’" (2023) https://www.euronews.com/green/2023/03/29/gravity-batteries-...
*** Special AI Seminar (abstract)
It has been widely recognized that AI programs require expert knowledge in order to perform well in complex domains. But knowledge alone is not sufficient for some applications; wisdom is needed as well. Accordingly, we have developed a new approach to artificial intelligence which we call "wisdom engineering". As a test of our ideas, we have written IMMANUEL, a wisdom based system for the task domain of western philosophical thought. IMMANUEL was supplied initially with 200 wisdom units which contained wisdom about such elementary concepts as mind, matter, being, nothingness, and so forth. IMMANUEL was then allowed to run freely, guided by the heuristic rules contained in its heterarchically organized meta wisdom base. IMMANUEL succeeded in rediscovering most of the important philosophical ideas developed in western culture over the course of the last 25 centuries, including those underlying Plato's theory of government, Kant's metaphysics, Nietzsche's theory of value, and Husserl's phenomenology. In this seminar, we will describe IMMANUEL's achievements and internal architecture. We will also briefly discuss our recent efforts to apply wisdom engineering to oil exploration.
How do you define "benefit"?
Your dad swindled thousands of "investors" out of their retirements and left you millions. You are benefiting from this and the children of the "investors" are suffering.
Your great-grand-dad swindled thousands of "investors" out of their retirements and you inherit a business empire. You are benefiting from this and the hundreds of great-grand-children of the "investors" are also suffering. They could've had inheritances but they didn't and work at Walmart.
You can trace your lineage to Thomas Jefferson who apparently owned 600 slaves over his lifetime. You still benefit from him having been a president and a wealthy man. You should have to trace ancestry of those slaves and compensate their current living family members.
But if your wealth comes from a line of crime, then yes, compensation would be adequate.
Practically, that's difficult. If you grew up wealthy because of generational crime that provides life advantages you can't return. At best, you could make sure you direct any inheritance to victims if possible or a suitable charity (and not your family foundation).
Legally, this is not plausible. All sorts of legal principles dictate that lawsuits must be timely (for various values of timely) and estates become unlitigatable not very long after they're closed. There are some cases in the news about crimes in WWII and such, though.
But the law knows such things in principle, even though usually not individually, but rather collectivly.
Like the native americans get some sort of privilege today. And (some) black americans demand reparations for past slavery.
But where to draw the line indeed. I don't think there is a universal answer.
Not tap unused or forgotten wells. This is purely risk avoidance, which usually means it won't get much attention or funding.
Leaky wells are a legal and insurance liability, which has a downstream impact on the financing of a drilling project.
In my opinion, that is like not fixing roads until someone collects data on potholes and forces you to, instead of actually keeping an eye on roads and bridges. A very American POV I'd say.
Of course, the company has to still exist
The system in place in the US means they mostly do not. A fund holding the amount of money it would take to clean up whatever you do on the land should be mandatory. Leave the land as — or better — than you bought it. Of course, that's very un-American.
Of course, that has not always been the case, and things falls through the cracks, but I would not immediately dismiss the entire industry as being non-compliant. I would dismiss the entire industry as flawed and needing change, but not on this specific point - it is vastly improved over past decades.
There's a point in time where this changed and permits needed at least a plausible expectation of remediation. If I had to guess that would have been late 1980s to mid 90s.
Most of the sites abandoned without remediation are from permits obtained before that time. I'm sure there's some cases where there was a setaside for remediation and it wasn't sufficient and the corporate entities involved went bankrupt, so it wasn't finished; but IMHO, most of the problem is older sites. Older sites also tend to have worse records, so there's that too.
Seems this is just a natural progression...neat.
Who led the adoption of GPT-1/2 there - a developer, you, a VP, someone else?
Another AI project I worked on there was a chemical tank use estimation and refuel application which used tank sensors, previous use history and some other metrics to pre-purchase and deliver product to keep tank reserves above a certain threshold.
For context, all of this was circa 2018-2021.
> downstream QE efforts
What the article describes sounds like it could have been built with 10 year old image processing tools and basic algorithms crunching the large amount of sensor data used to identify potential wells.
What makes this tool AI rather than an algorithm? Or machine learning?
That is very much manual intelligence
I mean, sure, these are methods broadly in the computer vision realm and that gets referred to as "AI" sometimes. But at the end of the day, this is "find all unfilled black circles of a specified diameter on these images". It's amenable to (and has been done by) traditional computer vision methods for a long time. There are certainly a lot of cases where a CNN type approach can perform better than traditional computer vision and there are always improvements to make.
However, I think it's a bit odd to treat this type of use case as some sort of AI breakthrough that wasn't possible or wasn't frequently done in the past.
Why can't normal standard work have a press release? Why do we need to play pretend and add buzzwords just to make things sound "cool"?
...But that's just me being a bit bitter, perhaps...
AI is useful for searching for targeted stuff where you can replace a person doing something that is probably pretty easy, but there is a lot of work that can be automated. Like searching for new viruses. AI has made identifying new viruses relatively easy and much quicker than a person, who typically tweaks input and data looking through what is noise to identify genome sequence of a new virus.
Were you complaining as heavily about OCR or Markov chains ever being referenced as AI in their hay day?
The term “AI” is in an infinite treadmill and the day it stops being useable as a time sensitive reference is probably the day it surpasses humanity and becomes its own State
LLMs aren't truly intelligent. [No True Scotsman fallacy...] They don't really reason. [A distinction asserted without giving a falsifiable definition of reasoning...] They're just next token predictors! [Which must be mutually exclusive with intelligence, I suppose?] Etc, etc, etc. Find your favorite pretext to dismiss modern AI, ignore the holes in the argument, and satisfyingly conclude that it's all smoke and mirrors.
Consequently you see hilarious takes from skeptics, like comparing today's enormous investment in AI to when people sold blockchain cartoon monkeys. Or claiming that modern models aren't useful for anything, as if they exist in an alternative reality where hundreds of million of people don't use them daily, and there's no incessant firehose of new tools/products/results discussed in news/social media constantly.
Folks won't let you use the right tool for the job anymore unless you make wildly hyperbolic claims about how groundbreaking it is and claim it's cutting edge AI.
The situation is bad for everyone. There's nothing wrong with using the right tool for the job and accurately describing it. I'm tired of having to inaccurately describe methods to be allowed to use them. E.g. claiming a Hough transform is "deep learning" so folks won't immediately dismiss it and demand I use some completely incorrect approach to a simple problem.
Classic computer vision is an utter PITA - especially when dealing with multiple libraries because everyone insists on using a different bit/byte order, pixel alignment, row/col padding, "where is 0/0 coordinate located and in which directions do the axes grow" and whatnot.
The modern "AI" stuff in contrast can be done by a human in natural language, with no prior experience in coding required.
It is true that all of this, machine learning, large language models, natural language processing and much more is AI, in the sense that it falls under the same artificial intelligence umbrella in computer science. It just feels a little like some one is using the term "construction" over and over, but what they are specifically talking about is some very specialized type of carpentry. It's not wrong, it's just not all that precise and give the wrong impression.
It would be really helpful if they called out how many potential wells were inspected and couldn't be verified. Are they confirming at 100% or 10%?
You can come across the land parcel claims sometimes by families that don’t have the capital to mine
[1] https://www.antipope.org/charlie/blog-static/fiction/acceler...