Saw an unexpected spike this night

Mercator wrote:

I will split the answer into two (I should have done that in my earlyer post) My detector detects mostly gamma's. it sits outside in a plastic casing to protect it from the invironment. Particles from solarflares (if any because of the atmosphere) are not likely to reach the tube. Gamma's however from solar flares or even supernovae should be able to reach the detector. not hindered by the atmosphere. So I presume it is possible to see 'gamma spikes'.
I wonder if this theory is plausible.

Apologies for the late reply. I have been on somewhat of a sabatical.

I'll try and break it down a bit because you are mainly correct, with caveats.

My detector detects mostly gamma's. it sits outside in a plastic casing to protect it from the invironment. Particles from solarflares (if any because of the atmosphere) are not likely to reach the tube.

Agreed. Most particles are blocked by the earth's atmosphere. You have to go up really high in order to detect them directly. Weather balloon, spacecraft, maybe a high altitude aircraft. So generally particles from solar flares are not detected on earth.

Gamma's however from solar flares or even supernovae should be able to reach the detector. not hindered by the atmosphere. So I presume it is possible to see 'gamma spikes'.

Yes and no. Solar flares produce very energetic gamma rays in the 10's of MeV. The earth's atmosphere blocks most gamma rays below 10~30 MeV so only the extremely energetic gamma rays get through. It would take one hell of a solar flare to give a noticeable spike. Similar with supernovae - it would have to be close to earth to have any measurable effect.

You may get the odd small fast spike every now and then, but would be hard to discern whether it came from the earth or space. Only those really massive solar flares would really be detectable and that also depends on it's direction and if you are in the firing line. They happen, infrequently, as far as I understand.

Mercator wrote:

What I meant with AI looking for paterns. I was thinking more that AI could find paterns in the movement of the radiation spikes over a large area. It could find spikes in a certain area an follow it with the wind over a landmass. detected by one station after another. Radmon will have a big database with CPM values and geographic locations.

Famous Dutch proverb: One fool can ask more questions then ten wise men can answer.

It can be done but would consume a large amount of development time, a very large amount. I asked ChatGPT what it would take to accomplish such a task (as that what it is good at) and despite the development time of writing and testing the api, to run the dataset (~30Gb) through it in raw form would cost around $241,592. (yes, that's two hundred and forty one thousand dollars, US.)

Step 1: Convert bytes → tokens
1, Your dataset is 30 GB raw. That’s 30 × 1024³ bytes ≈ 32,212,254,720 bytes.
2, Each character is roughly 1 byte. But GPT counts tokens, not bytes.
 - Approximate rule of thumb: 1 token ≈ 4 characters for English text.
 - So 30 GB → 32,212,254,720 ÷ 4 ≈ 8,053,063,680 tokens (~8 billion tokens).
 
Step 2: Token costs
 - Using OpenAI’s pricing for GPT-4-turbo (most capable large model for this task):
    - $0.03 per 1,000 prompt tokens (for simplicity, using “prompt” tokens only).
 - 8,053,063,680 tokens ÷ 1,000 × $0.03 ≈ $241,592

So in order to get this cost down to something more manageable the data would need to be pre-processed, summarized, and/or converted into structured prompts or embeddings before firing at the API. I'm not sure I would be capable of programming such. Maybe, but I have only used the API in a small way with PHP. This would require several different processes to process the data prior, so may be a little more complex than I could manage. In doing so though it could get the costs down to somewhere between $100 - $1000 per run. And of course, once historic data is processed, it doesn't need processing again, only the later part, maybe a month of history and 'today'. The the next run would be one day less history as a new day is added, or week, or month, whatever, so after the initial run to process history, if we even did that, each subsequent run would be much cheaper, maybe in the order of $10-$50 dollars per run.

So after a huge amount of development and testing to implement it, it would still incur a running cost of approximately $10-$50 per run.

Another thing is the pre-processing of the data would have to be done on a fast multi-threaded computer for it to be done in good time. The current server just couldn't handle it at all and would bring it to a crawl for hours.

So, yes, it could be done, but at great development time and then ongoing running costs. 

To add to this; if this were to run live, or an almost live fashion, it would be chewing up that $10-$50 every few minutes.