Thanks for the suggestions. A small package containing 4.5Kg of lead arrived by post today. "Be careful, it's heavier than it looks" said the young lady postie as she handed it over. "That's what I was hoping", I replied. And left her to wonder what on earth she had just delivered...

Steve

Then, as I already suspected, you are a far better photographer than me!!

Steve

No. But I'm a scientist who drifted into engineering (and then into retirement ) and along with some colleagues had started to think about some of the implications of AI for safety engineering. I now try to keep up with technical developments in AI by reading various 'public understanding of science' sources. I've also been an amateur musician throughout my life, still performing a few times a year (and not making any money from it!). And it's the live performance aspect of music which makes me confident that AI will not take over music. The human element is what people want, either from live performers or from a DJ who is choosing and playing recorded music. Who wants to go to a gig where all you hear is AI generated imitations blasting out from a speaker?

And now for something completely different: should we be worried about AI because it learns? We all benefit from having various metering modes on our camera. Sometime, the chosen metering mode doesn't give us what we want. Suppose there was an AI chip in the camera which could learn whether the exposure from the chosen metering mode would be what you really wanted then apply exposure compensation to get the exposure you actually want. Here's how it works. After processing photos, you tell the chip whether each exposure was OK or how much it needed to be changed (ideally automatically by wifi or Bluetooth directly from post-processing software to camera next time both are on). The chip works out how much exposure compensation you are wanting for each of the types of photo you take and can use that in future to improve exposures for your photos. Sounds useful to me. But it could only be achieved by AI 'learning' what you want, and what it 'learns' would be different for someone else. What's wrong with that? Definitely nothing to be worried about IMHO. In fact, I'd rather like to see it in the next round of updates to Pentax DSLRs!

Steve

Not true. AI is also based on patterns or algorithms. In essence, AI tools are enormous curve-fitting algorithms, no different in principle to the curve fitting that I learned at school in A-level maths (is it now in the GCSE syllabus?). Given some data points (e.g. distance to object and AF accuracy at that distance), find the best-fit curve then use that curve to predict the values of new data points (e.g. AF accuracy at distances not in the original data set). The differences are in size, scale and scope. AI tools have millions (or more) of parameters which are tuned (via 'training') to give a best fit to the data (e.g. photos scraped from the internet). But they are hellishly complicated and in most cases the way they curves are fitted, the 'best fit' parameters and the quality of the resulting fit are so deeply hidden that nobody yet knows what they are. However, like all curve-fitting algorithms, outputs are always more reliable if they fall within the range of the data used to fit the curve, weird things can happen well outside that range. And that's why AI can produce strange things ('hallucinations') and why AI always wants more and more, and more varied, training data.

So did the motor car and computers. Yes, many jobs disappeared. But new types of job emerged. How many motor mechanics or computer programmers were there in the 19th century??

Lots more I could write, but lunch is calling...

Steve

There are 2 strands in what you say. So here are some thoughts.

The first strand is people being deprived of their jobs. Well, in the 19th century there were vast numbers of people copying documents by hand using pen and ink. Then the typewriter came along and those jobs were lost. When I started work (many years ago) there was a typing pool where skilled typists turned nearly illegible, handwritten manuscripts into neatly typed reports. Then the word processor came along and the typing pool vanished. I could also mention the agricultural labourers, saddlers and straw hat makers who appear in my 19th century ancestors. Not many of those jobs around today thanks to mechanisation and the automobile. But other jobs appear and fill the gap. At the personal level, it's hard for people who lose their jobs for whatever reason (and it's happened to me, too). However, jobs have always come and gone as technology changes. AI is just another turn of the wheel.

The second strand is the use of one's work to train AI. There is real substance here, but I do wonder whether the creative community (including photographers) are aiming at the wrong target here. Creative people have always learned from other people's work, then gone and produced works of their own which are derived, in part, from the works which they have learned from. Isn't that one of the reasons that forums such as this exist? It seems to me that the real problem is not learning from other people's work but whether the new work that is produced is sufficiently similar that it breaches copyright. The 'red bus on Westminster bridge' case (link) established that a photo can breach copyright even if it is about common subjects and details differ significantly. The judge stated "On the question of copying, I find that the common elements between the defendants' work and the claimant's work are causally related. In other words, they have been copied." So suppose an AI thingy which learned from material scraped from the internet were to produce, say, an astro image which had foreground that looked similar to a Spad masterpiece. It seems to me that there is a clear argument for breach of copyright unless it were possible for the AI folk to show that the thingy had not used Spad's masterpiece in its training. I look forward to the first such case appearing in court!

Steve

I'll have the Huckleberry's sausage & egg hot sandwich, please, followed by a large castle bucket set .

Agree 100% .

Steve

It's almost obligatory for a forum member to be "getting on a bit". Most younger folk use Facegram, XTok or similar and are happy to be influenced by influencers, advertisers and worse, even if they think they can resist the 'fluences. But once you reach "getting on a bit", you find yourself wanting the civilised discussion of a forum. That's my theory, anyway. And I worked it out without any 'fluence whatsoever.

Steve

Who on earth would take one of the greatest portrait lenses of all time and put a Raynox close-up attachment on it? Well, I have both so why not try it... Very thin depth of field, of course. Ideal for experimentation. All with a K3iii.

Thoughts, comments, cries of 'sacrilege' welcome, as usual.

1.



2.



3.



4.




Steve

Dive weights - brilliant! I was wondering about gazebo weights - thought there were some in the garage from years ago but can't find them. What I'd really like is a big block of platinum, but the budget won't stretch to that...

Steve

Plants, flowers, bits of plants and flowers and anything that takes my fancy. I've even done dried paint on a stirrer. Lichen is on the agenda. But I don't do dead insects, eyes of flies, that sort of thing.

Steve