I've played with ChatGPT and am very unimpressed with it. The answers it gave were shallow and lacking any indication of understanding the concepts behind the question. They were also repetitious, with information regurgitated in different ways.
With regards to AI and astrophotography, I'd much rather struggle and battle my way manually through the process and achieve the satisfaction of creating a good image.
For those of us that have gone through the Astro learning and investment curve, i suspect many are as in love with the process as much as the outcome. It is a ritual that we find satisfying. And we spend a lot of time talking here about process.
I'm 3 years into this hobby now (but have been keeping an eye on it for about 12 years), and I hated that learning curve. I still do.
So much of this hobby is rooted in antiquated process and technology.
So I've been jumping at the change to try the New Things (tm) that come out. Whether that's in software (NoiseX, Starnett, StarX, BlurX, StarTools), hardware (ASI AIR, AM5), scopes (more petzvals please!) it's all good. Because the hobby has been slow to change, but these past few years when younger people have entered the market, we've seen a massive explosion in new ways of capturing images.
I guess it also depends on where you're coming from... I'm not a data scientist. I don't actually care if my image suddenly has 5 pixels extra of hydrogen filament created by software. The people I share these images with outside of the astro community don't care either. I'll leave the pixel peeping for NASA and astronomers and scientists who need that sort of accuracy.
But if you're the sort of person who wants data integrity as the first port of call, that's great. Sadly these AI tools are probably not for you. Thankfully the old processes are still there if you want them.
I've played with ChatGPT and am very unimpressed with it. The answers it gave were shallow and lacking any indication of understanding the concepts behind the question. They were also repetitious, with information regurgitated in different ways.
This "AI" is not more than fancy pattern matching
ie: these phrases are associated with these other phrases. The filtered phrases are passed on to another pattern matcher that makes reasonable sentences. Done.
Location: '34 South' Young Hilltops LGA, Australia
Posts: 1,482
AI assisted eclipse photography
I have been working with Terry Cuttle on a series of articles about solar eclipses and the 5 Australian eclipses for the AAQ website.
I have mostly been doing a lot of proofing and editing of Terry's articles. Terry asked me to write a 3-page article on eclipse photography. Easy I thought. Well no as it turns out. I ran into what you might call "reverse writers' block," or perhaps, "writers' verbose diarrhoea."
My first draft was a whopping 17 pages. I viciously edited it down and I only got it down to 9 pages then I stalled and could cut no more.
So, I asked ChatGPT to write the article for me. So, I asked, "How do I photograph a total solar eclipse?"
Interesting exercise. It produced a succinct, fairly well-written, 1-page article. However, it was riddled with factual errors, omissions of critical info, and bad advice. It made an excellent starting point for me to correct the errors, omissions, and bad advice, and then turn it into the short article I needed to produce.
It basically told you to use a solar filter then gave instructions how to underexpose the partial eclipse by about 5 stops. Then it failed to tell you to remove the filter for totality, just to “check the live view and adjust the exposure settings accordingly.” Hmmm, with a solar filter left in place, you need a 16 minute exposure to record the corona.
I corrected the errors, fixed the most egregious omissions, then added some of my diagrams and a few extra bits of advice and voila, I had a 4-page article.
There is a lot of debate in the education sector about whether students will be able to get away with cheating using these tools. From what I have seen, the accuracy is about as accurate as a feature written by one of the writers on space.com. For a longer student assignment on a subject like history, if the error ratio is as bad as the drivel I got out of it, teachers have nothing to worry about. It is nonetheless a very useful tool in the right hands and used with suitably expert human oversight.
I am including a screen shot here of the original ChatGPT article draft with my comments for your amusement:
I have been working with Terry Cuttle on a series of articles about solar eclipses and the 5 Australian eclipses for the AAQ website.
I have mostly been doing a lot of proofing and editing of Terry's articles. Terry asked me to write a 3-page article on eclipse photography. Easy I thought. Well no as it turns out. I ran into what you might call "reverse writers' block," or perhaps, "writers' verbose diarrhoea."
My first draft was a whopping 17 pages. I viciously edited it down and I only got it down to 9 pages then I stalled and could cut no more.
So, I asked ChatGPT to write the article for me. So, I asked, "How do I photograph a total solar eclipse?"
Interesting exercise. It produced a succinct, fairly well-written, 1-page article. However, it was riddled with factual errors, omissions of critical info, and bad advice. It made an excellent starting point for me to correct the errors, omissions, and bad advice, and then turn it into the short article I needed to produce.
It basically told you to use a solar filter then gave instructions how to underexpose the partial eclipse by about 5 stops. Then it failed to tell you to remove the filter for totality, just to “check the live view and adjust the exposure settings accordingly.” Hmmm, with a solar filter left in place, you need a 16 minute exposure to record the corona.
I corrected the errors, fixed the most egregious omissions, then added some of my diagrams and a few extra bits of advice and voila, I had a 4-page article.
There is a lot of debate in the education sector about whether students will be able to get away with cheating using these tools. From what I have seen, the accuracy is about as accurate as a feature written by one of the writers on space.com. For a longer student assignment on a subject like history, if the error ratio is as bad as the drivel I got out of it, teachers have nothing to worry about. It is nonetheless a very useful tool in the right hands and used with suitably expert human oversight.
I am including a screen shot here of the original ChatGPT article draft with my comments for your amusement:
Just wanted to follow up to make it clear that this earlier description of how you might think ChatGPT works is incorrect.
Quote:
Originally Posted by oska
This "AI" is not more than fancy pattern matching ie: these phrases are associated with these other phrases.
The filtered phrases are passed on to another pattern matcher that makes reasonable sentences. Done.
On the contrary, where it describes itself as a "state-of-the-art language
model, trained on vast amounts of text and equipped with deep learning
algorithms", these are not a set of buzzwords but instead the
precise technical terms used by computer scientists and engineers to
characterize this type of system.
Deep learning language models such as ChatGPT are neural networks
using a probabilistic approach to derive the output sequence from the
input sequence.
By comparison, pattern matching systems in the way you describe are
relatively simple algorithms.
And the punctuation of 'Done' as a kind of QED belies the fact that
this system is vastly more complex and sophisticated than you might
appreciate. This system is not just large, how it works is non-trivial.
In fact the description you provided showcases one of the challenges for
training systems such as these. They reply on a large corpus of text for
their training and that overwhelming came from the web.
And of course most of the web is written by humans. So the old saying
of GIGO - garbage in, garbage out - is absolutely true. I am trying to
coin this in the politest way without offending, but the description you
provided might be a good showcase for garbage. If a language model
were to parse it, it would add to its number of incorrect facts. It is a tough
problem, hence ChatGPT was trained on sources that tended to have
a higher probability of being factually correct compared to say opinions
on a chat forum.
As an engineer and a member of the Institute of Electrical and Electronics Engineers
and its associated Computer Society for 45 years, I've watched AI systems evolve over time.
I've been posting on these systems in the Science section since June of
2021 and you can follow one of those threads there if you would like to
learn more. :- https://www.iceinspace.com.au/forum/...d.php?t=192039
Garry, I said fancy pattern matching. Obviously that's not simple grep or even google's search thingy, that's just silly.
In its barest engineering essence without using any hype, buzzwords, overly technical terms or overstating anything, in as few words as possible what is it technically then?
What are "language models" and "deep learning algorithms" if not a bunch of weighted links on how phrases relate and fancy pattern matchers?
For your GIGO argument to be valid you would need to show the source being wrong in a statistically relevant manner ie: the obvious mistakes it makes with respect to context, relevance and accuracy being predominant in the training material that was used to train it, such that it "learned wrong". To me the "Garbage Out" nature of its responses specifically and in general demonstrate it being merely a fancy pattern matcher, a "next gen google" if you must, rather than anything that would reasonably be called even rudimentary AI. Sure, better training will help mitigate some of the short comings but it will not change what it actually is.
I think it's wildly fascinating, especially the black box nature of the resulting networks.
I have been involved in astronomy since I was 9 and I am now about to turn 59. In that time much has changed. More and better equipment available to amateurs. No need to make everything yourself anymore. It's gone from photography (using dark room techniques) to imaging (using software techniques). The learning is part of the fun and satisfaction. This hobby is not for the impatient!
Along the way there has been a constant learning curve. Whilst it can seem like it is now plug and play it's not in reality. You will need to have a mechanical aptitude to setup, service and repair equipment. You still need to problem solve on occasions. Processing data still requires an element of skill and understanding of the science of Astronomy. Processing is about 50% of the exercise. If you think you just plug it in, walk away and then see the finished result later, you're dreaming.
As to AI sharpening, noise reduction etc I don't agree that putting garbage in you get silk out. You need good weather conditions, good setups, good guiding etc. I see AI deconvolution and noise reduction as tools only. Tools in an already very large tool box. I like many experienced astrophotographers use several programmes when processing. Some elements of processing just require the use of photoshop every single time. A few tedious aspects need automation. AI can't give me altitude based images, but it can improve the point spread function of the data and reduce the noise levels a little, but I still need to take a lot of data, it still has to be good data. That though does not make great images. There is a lot to do in processing aside from these tools. There is still a strong element of artistic leaning and understanding chemistry and knowing what looks natural when processing. There is still the need to sit and do the work to produce the results. I imagine this will be the case for some time. When it becomes purely automated, then what will be the point of doing it. The truly lazy will love it I imagine but those of us that love the maker space will still doing things the "hard" way.
We are at an interesting moment in time - where a small freeware version of an AI has caused massive stir and interest in A.I. in the public domain. I highlight small freeware version in comparison to what large Governments, Military, Mega corporations and well funded intelligence agencies can use at their discretion. One wonders if ChatGPT is a mere demonstrative toy compared to their capabilities.
I do see the potential this has to reshape modern life - in many aspects. A human being will be defended in USA for the first time this month by an A.I. Its ability to code and improve code vastly outstrips many human developers. It has already passed (bottom 20% of passes) the final law exam for Harvard law school graduates and it is growing more sophisticated at an accelerating rate.
Whilst it isn't Skynet - one wonders what will happen when AI's start feeding themselves and designing better and better AIs - the way organic creatures evolve. It isn't aware software with its own intelligence - it is meant to replicate behaviour or responses in a way to statistically acting in accordance as if it were. So outside its black box it appears frighteningly capable, inside the box its data and algorithms that are whirling around executing their models - and this will likely continue until someone tells the world they have invented a quantum level AI.
I have no way of knowing where this will all lead - the same way I have never seen a zombie apocalypse or end of the world movie were we all ran out of toilet paper - but hey who during Covid foresaw that?
I do foresee a time and real soon when made by AI surpasses any made by human hand or skill - and it actively gets marketed as such. I sense it will be a rather wild ride ahead. I wonder how far we are from the day you can ask an AI "What could you design and built that would supplant and replace you with something far more advanced?" and we get to see what an AI thinks will make AI itself obsolete...
My son and I were discussing AI going forward, and the big leap that GDP-4 will bring. Imagine a fleet of AI enabled Starships leaving for Mars, onboard are a troop of Boston Dynamic AI robots, tasked with assembling the Boring machine, 3D habitats printers etc. These hive minded robots running around on the surface, busy with their tasks, with no risk to any human. They return to their charging station when required. Sounds like the Borg to me.
AI - An existential risk to humanity if its rational capacities are not steered towards beneficial goals.
Yes, as a visual astronomer, I wear the Luddite mantle with pride, and reflect with considerable dismay the forthcoming possible lack of ability in all the basics by coming generations.
Maybe I'm just a cranky old man who was caned at school after being caught with a Texas Instruments calculator.
AI - An existential risk to humanity if its rational capacities are not steered towards beneficial goals.
Yes, as a visual astronomer, I wear the Luddite mantle with pride, and reflect with considerable dismay the forthcoming possible lack of ability in all the basics by coming generations.
Maybe I'm just a cranky old man who was caned at school after being caught with a Texas Instruments calculator.
Only a cane? Heck we got the cat of nine tails with two extra tails if we were caught with an abacus.
Control by fear of hurt is just wrong.
And thinking back what miserable people were those who bullied children with the cane.
AI is actually real as evidenced by the fact it is being discussed as well as other things establishing it as very real indeed.
Hardly bogus but perhaps you could describe how that could be.
Counterfeit involves a copy passed off as the original and clearly this has not happened.
Ersatz how exactly..meanwhike I will look that word up...
Fabricated..well thank goodness as conjured out of thin air would be of great concern.
Faked..then there is nothing to worry about as the claims will come to nothing.
False..again if false there is nothing to worry about.
A list of words is hardley a reasonable substitution for a well reasoned arguement and smacks of pulling on the emotions without any basis for the general misgivings.
AI is good ( let's start from there) and the negative responses come mainly from folk who will even feel threatened in the presence of a human who has a better education and a wider life experience...
There is nothing to be proud of being a luddite... these were a group of people who damaged property in a silly attempt to turn back the clock...shameful read about what took place.
Moreover why can't folk see just how these things will eventually move the human race forward because all this says to me is we are acquiring better tools to manage...tools are what have made us great ... I can just hear some old cave man growling when his son showed him the first flint spears head.." no no no this will ruin everything etc etc"
Even a casual understanding of evolution makes one realise at least one important characteristic of a species surviving is it's ability to manage change... so why not look at the the benefits and look forward to when you will be a beneficiary.
Can't we as old men we set an example to those who come behind us that seeing the best in things before us it the way to be, that making praise our goto is better than being negative....I have found in general people like to be in the company of someone who is positive and hopeful so I would think it is a sensible choice....still no statues of a critic that I know of...
Only a cane? Heck we got the cat of nine tails with two extra tails if we were caught with an abacus.
Control by fear of hurt is just wrong.
And thinking back what miserable people were those who bullied children with the cane.
AI is actually real as evidenced by the fact it is being discussed as well as other things establishing it as very real indeed.
Hardly bogus but perhaps you could describe how that could be.
Counterfeit involves a copy passed off as the original and clearly this has not happened.
Ersatz how exactly..meanwhike I will look that word up...
Fabricated..well thank goodness as conjured out of thin air would be of great concern.
Faked..then there is nothing to worry about as the claims will come to nothing.
False..again if false there is nothing to worry about.
A list of words is hardley a reasonable substitution for a well reasoned arguement and smacks of pulling on the emotions without any basis for the general misgivings.
AI is good ( let's start from there) and the negative responses come mainly from folk who will even feel threatened in the presence of a human who has a better education and a wider life experience...
There is nothing to be proud of being a luddite... these were a group of people who damaged property in a silly attempt to turn back the clock...shameful read about what took place.
Moreover why can't folk see just how these things will eventually move the human race forward because all this says to me is we are acquiring better tools to manage...tools are what have made us great ... I can just hear some old cave man growling when his son showed him the first flint spears head.." no no no this will ruin everything etc etc"
Even a casual understanding of evolution makes one realise at least one important characteristic of a species surviving is it's ability to manage change... so why not look at the the benefits and look forward to when you will be a beneficiary.
Can't we as old men we set an example to those who come behind us that seeing the best in things before us it the way to be, that making praise our goto is better than being negative....I have found in general people like to be in the company of someone who is positive and hopeful so I would think it is a sensible choice....still no statues of a critic that I know of...
Alex
Going with the insults again rather than debating the point. Just because a person that has reservations about AI doesn't mean they are less educated or lack life experiences. The reality is AI is here whether it is good or bad time will tell.
Going with the insults again rather than debating the point. Just because a person that has reservations about AI doesn't mean they are less educated or lack life experiences. The reality is AI is here whether it is good or bad time will tell.
I was not attempting to insult so I appologise for the way I put that....just here alone desperately craving to be heard... actually I was trying to be humorous which was silly because that never works...anyways just because I say what I said about education and experience does not mean anyone here has to identify as that... most of all Hans I am sorry that you find me so annoying...well I am not sorry I suppose but it seemed like a nice thing to say...I can live with the fact many folk think the same way about me as you do..Perhaps I should care but I don't folk can think what they like which was really the point you were making I guess.
I find all the negative talk that surfaces here tiresome and that's my excuse for my bad behaviour...
