I was trying to use gemini 2.5 flash image / nano banana to tidy up a picture of my messy kitchen. It failed horribly on my first attempt. I was quite surprised how much trouble it had with this simple task (similar to cleaning up the street in the post). On my second attempt I had it first analyze the image to point out all the items that clutter the space, and then on a second prompt had it remove all those items. That worked much better, showing how important prompt engineering is.
That actually proves how important the “number of attempts” metric is. It’s not just a “make everything pretty” button - it’s more like a powerful but slightly dumb intern who needs clear, step-by-step instructions. Your two-step approach really captures the essence of prompt engineering
Yeah, that's part of the reason I list the number of attempts as part of the stats for each model + respective prompt. It's a loose metric of how "steerable" a given model is, or put another way, how much I had to fight with it before we were able to get it to follow the prompt directives.
The wikipedia article you linked mentioned that the Goiânia source was 50.9 TBq (1,380 Ci) when lost, the source in question here is probably 1-10 mCi, so about a million times weaker. See this other thread about the incidence:
I am a bit surprised they haven't found it a >1mCi Cs137 source should be quite easy to find, it can be detected with a medium sized detector from >10m distance. A car driving along the road with a bunch of detectors in the trunk or a helicopter with a stack of NaI bars mounted on the bottom should see the radiation emitted by the source.
I am working at a National Lab in the US and this is the main focus of my research. Please have a look at our website if you are interested in learning more about this type of things [1].
Nice, so, just speculating with what is in the public domain here;
* what kind of counts would you expect from a 19-becquerel caesium 137 ceramic source in a 50 litre NaI spectrometer ~ 40 metres away in a 1 second window ?
I mean the spectrum is going to look like [1], of course, but I'm imagining a crop duster [2] draping at 70 metres / second for along the centre line of the 1,400km stretch of road, maintaining a 40 m clearance and coming preloaded with the 256 NASVD Hovgaard kernel of the West Australian Outback looking for a Cs137 peak and compton.
Would you be expecting any radon interference here?
Of course the fun starts if our nugget of interest isn't on the road but was picked up in a 4x4 | truck tire and carried along for a few hundred km.
Would you expect a contact trace to show up along the road here, good enough that if it dies off one can investigate whether it fell out, or got carried elsewhere by the vehicle turning away?
19 Becquerel is quite weak. My back of the envelop calculations of the source mentioned in this article makes me guess it is about 1 milli Curie = 0.001 Curie = 37 Million Becquerel.
For simplicity lets assume 50 Liter is a 1 meter x 1 meter x 5 centimeter thick panel. 5 centimeter is usually sufficient to give you a good change to measure gamma rays at most energies and you want to have a large surface area exposed to the source, so this seems a quite optimal configuration for 50 Liter of NaI.
A source emits gamma rays in all directions uniformly, so to calculate the fraction of gamma rays that end up in our detector we calculate in "solid angle space".
The surface area of such the detector is 1 meter^2. The solid angle coverage can be approximated by surface area / distance^2 [1], so at 40 meter that would be about 0.000625. The full solid sphere is 4PI ~ 12.5, which is the surface area of a ball with 1 meter radius.
The fraction of the full solid angle that is covered by the detector is 0.000625 / 12.5 = 0.00005.
You have 19 Becequerel = 19 Gamma rays per seconds leaving the source so 0.00005 * 19 Bequerel * 60 seconds = 0.057 gamma rays per minute reaching the detector. You won't see this over the background.
Putting 37 Million Becquerel into this equation, we will get roughly 111,000 counts / minute in the detector. Remembering correctly (and this is only a rough guess) you would expect about 700,000 gamma rays per minute from background (what we call naturally occurring radioactive materials or NORM) in that much detector volume. So that is about a 1:7 signal to background ratio. You should have a very good chance of seeing that.
I neglected here that not all gamma rays that reach the detector actually will leave a signal in the detector. But at 662 keV, the energy most of the gamma rays in Cs137 are emitted at, about half of them do so in a 3 cm thick detector. At 5 cm I would estimate about 75% or more doing so.
Hope that gives you an idea how you can estimate these types of questions.
Edit: Forgot to mention the radon. Usually you do not measure much radon unless it rains. It is mostly a dessert there so I don't think that should be an issue
If cosama doesn't make it back, the wording here is generally taken that "source" is the source material that radiates alpha, beta, gamma radiation and is in some sense 'pure'.
As such it radiates equally in all directions.
Of course a manufacted Source(TM) for industry might actually be a slug of doped Cs-137 surrounded by a lead casing with a window to direct emissions.
This still leaves us talking about the actual source of the radiation which emits equally in all directions .. but being blocked in some directions by either a casing or perhaps the earth below, a pool of water, etc.
Thanks for the answer. You are absolutely correct, although once you start taking all these effects into account it is often easier to run simulations. For a back-of-the-envelop calculation assuming uniformity is quite common.
I'm glad you picked that up as 1.9 MBq through up to 750 MBq is more of the range I would have expected for a mining instrumentation source [1].
I'm guessing both the The Guardian [2] and AAP [3] relied on the same reporter who missed fact checking and an M (and possibly a decimal place) when interviewing the W.Australian Health minister (who, of course, may have bumbled that himself).
The W.Australian official press release avoided details [4].
Take note when going forward in your career as there'll likely be some moment of unexpected public relations in the future.
Any thoughts on the "full quote" ?
Chief Health Officer Andrew Robertson said the small silver cylinder was a 19-becquerel caesium 137 ceramic source commonly used in radiation gauges.
“That may not mean a lot to people but probably more concerning is that it does emit a reasonable amount of radiation,” he said.
Dr Robertson said the unit emits about two millisieverts of radiation per hour, which is the equivalent of having 10 x-rays in an hour.
“Two millisieverts is also the amount of natural radiation we would receive in a year just by walking around,” he said.
“This is a source we have to be very careful of … It is quite a large radiation dose.”
It's one thing to drop an M in MBq, another to get x-ray equivilants incorrect, so does this sound about ballpark to you?
I confess to a distaste for Sieverts as a per kilo absorbed dose (or is that Grays) as its kind of subjective to target rather than source properties (as the source may be partially shielded by lead, water, etc.)
Moving on, 700 k counts / minute background (or 12 k counts/sec for those of us working in one second over lapped accumulation windows) sounds ballpark.
The interesting twist on the source, though, is it might be shaped in the sense of having a wrapper of lead (say) with an aperture of preference to direct the radiation preferentially.
> I neglected here that not all gamma rays that reach the detector actually will leave a signal in the detector.
For at least two reasons, 'eh?
* Not all gammas passing through the crystal will cause a flash (you covered by implication), AND
* Not all flashes will be counted (oft neglected) as the scintillation counters get saturated by single events and need to recover, they also fluff damn near simultaneous multiple events as singles (all this dependant on scintillation detection equipment).
> Hope that gives you an idea how you can estimate these types of questions.
I was very interested to hear your reasoning, I appreciate it and I hope your comment gets a few eyeballs, although HN moving on apace as it does, who knows?
> Forgot to mention the radon. Usually you do not measure much radon unless it rains. It is mostly a dessert there so I don't think that should be an issue
Hmm. You might want to look into this further I suspect.
My poor understanding is that rainstorm fronts drop the barometric pressure which draws radon gas from the cracks and crevasses in which the gas has already been expressed from the rock sources.
So, yes, there is an increase in radon gas where radon gas is normally expressed due to pressure drops.
However, in a hypothetical environment with no pressure changes .. radon continues to be expressed if it is being created.
I've found, in my humble experience, that even on clear days rain free days in Western Australia radon is thick on valley floors in those areas that have it in the still of the mornings, and gone later in the afternoon as the winds increase and clear the heavy ground hugging gases.
I would suggest that radon is a function more of geology and topography .. modified in variance by air pressure and wind conditions.
In any case, thank you for your response, I look forward to any further thoughts you may have, and, in closing, you may enjoy this entertaining A0 wall map [5] .. or perhaps the larger raw datasets (also available) [6]
Turns out the dropped a G not a M [1]. 0.5Ci is a very strong source, it should be observed by a plane at 100-200m above ground with sufficient detector volume. Considering they still haven't found it, it must have rolled somewhere, where there is considerable shielding upwards or towards the street, or been carried away.
@defrost Great discussion and addition by the way. I only have one additional comment concerning radon. Radon is produced in a radioactive decay in the ground, and as it is gaseous it usually dissipates into the atmosphere rather quickly, so there is not too much of it near the detector at any given time. As you mention correctly, certain weather condition obviously can trap it near the ground. We usually study urban centers, with lots of asphalt covering the ground, affecting the amount released near the ground and near our detectors. We usually see it during rain, as then the radon in the atmosphere is washed down through the rain and accumulates on top of the ground for a while, which we often see as a strong uptick followed by an exponential decay. I have a coworker that just submitted a paper for review regards modeling this, sorry can't link the work yet. Overall, radon is quite tricky to understand and what you mention seem very reasonable. It would be quite interesting putting a detector for a while in these areas. Nevertheless, the measured amount of radon usually fluctuates on larger scales, maybe hundreds of meters, while the source will create a very localized signal, so it shouldn't cause too much troubles for a search like this.
Yeah, mind you the Giga is quite a jump over Mega (see the reference I gave above on mining applications), so I'm not entirely sure I trust the update.
For context, Rio Tinto operations in Mt. Newman are mining iron ore from near surface open pit | mesa deposits; blasting rock with minimal crushing to fit in rail carriages for transport to the ocean ports where ore is crushed smaller, screened and graded (blended) for shipping overseas .. in total some 850 million tonne per annum of iron ore from that region (not just Newman alone).
I'm guessing the guage is for estimating ore density | grade on the fly as rocks tumble through the load out (from the remote operated semi autonomous near robot trucks) into kilometre+ long robot trains.
> Considering they still haven't found it
Yeah, I'm very surprised the search is still on, I'm leaning towards it's well off route in a tire in a truck (or already fallen free) OR some heavily tatted FiFo worker (fly in | fly out) nicked it as a souvenir for the local bikkie clubhouse and hasn't came home from a holiday in Thailand yet .. .
Radon: IIRC there are additions in the Australian AGSO radiometric processing guidelines re: radon removal that derived from work I did levelling out radon smearing back in the 1990s - somewhat out of date these days, but hey, that's the world for you.
There are also potential modifications to the work by Brian Minty | Jens Hovgaard | Bob Grasty using an NASVD to firstly highlight the use of Radon eigenvectors and then to rotate them out of the signal space in the processing pipeline.
I'm guessing (completely w/out context) that perhaps your friend is moving down that path (or perhaps well past and elsewhere, as I said, it's been a while).
If yourself, your friend, or anyone else gets bored searching the sidewalks for traces of terrorism (or whatever the hell it is that you're doing and not talking about) .. have a look into this:
A radioactive source that weak won't emit much heat. You can touch it and it will not be hotter than a piece of metal. You would need a source quite a bit stronger (at least thousand times stronger) to actually feel/see the heat from it.
In my opinion there is only one humidifier that is worth the money, the honeywell warm mist humidifier [1]. It works great for us, just put it in a little tray in case it leaks. It also got one of the best written customer reviews on amazon I ever saw [2]. TL;DR, they use it on McMurdo station in Antarctica and love it.
Don't buy a ultrasonic humidifiers, they actually reduce the air quality considerably [3], you need to get an evaporative humidifier and frankly there aren't that many.
