MJB
Well-known member
An early 'advance' in AI was the removal of watermarks... (Yes, initially, the results were crude, but now...)
MJB
-
Welcome to BirdForum, the internet's largest birding community with thousands of members from all over the world. The forums are dedicated to wild birds, birding, binoculars and equipment and all that goes with it.
Please register for an account to take part in the discussions in the forum, post your pictures in the gallery and more.
AI and rare birds (1 Viewer)
- Thread starter Mysticete
- Start date
MJB
Are these 1000s of people using AI? Or the product of AI? Confused
Xenospiza
Distracted
It is picture-recognition software.
Using AI to generate products and then using these products...
I've experimented with AI generated bird images and they are universally terrible. Just asking for a straight forward bird like "Robin" results in something that looks a bit like a Robin but the details are all wrong. If you ask for a "Robin on a bird table" the results can be comical.
One could generate Big-Foot or Loch Ness Monster quality images to try to fool the gullible but trying to string some serious twitchers would take a serious leap in capability.
One could generate Big-Foot or Loch Ness Monster quality images to try to fool the gullible but trying to string some serious twitchers would take a serious leap in capability.
Yes, that's what I was suggesting. Using AI to generate a lifetime of ebird records, and uploading, and uploading, and uploading!
Pretty soon the AI generated bulk uploads would rival the real data.
The data could be set to be in the reverse chronological order so every trend we have seen in ebird would be negated ! Every population would be static over time
Edit:
E.g.
1. Take all users in ebird with Hen Harrier sightings in North of England
2. Download their ebird data
3. Reverse the dates, throw in few woodpigeons and robins
4. Create new user and upload a new lifetime of birding sightings
5. Move to next user and repeat.
Result Hen Harriers stable population over the last 40 years in Northern England.
I could do this without AI, so you could only imagine the chaos you could unleash with AI to automate and randomise. It seems to be limited only by our imagination.
Last edited:
dantheman
Bah humbug
It would be an interesting experiment to say, produce a few images weekly or monthly and post them on here. See the advancement or not. (Different AI generators too perhaps). I'm sure people must be doing this, perhaps not with birds though.
Would be an interesting experiment. There are youtube videos of AI generated images with lots of views, not sure how much they are cherry-picked/curated though.
Hi Everyone,
I'm new to the forum, but I'm a big fan of birding and I'm interested in the potential impact of AI on the hobby. I saw the thread about AI-generated images of rare birds, and I wanted to share my thoughts.
I think it's definitely a possibility that AI-generated images could be used to prank or deceive people about rare birds. However, I also think that AI could be used to help birders in a number of ways. For example, AI could be used to develop new identification tools that are more accurate than current methods. AI could also be used to track bird populations and migration patterns, which could help us to better understand and protect these species.
Overall, I think the impact of AI on birding is still uncertain. However, I believe that it has the potential to be both positive and negative. It's important to be aware of the potential risks, but I'm also excited about the potential benefits that AI could bring to the hobby.
I'm looking forward to learning more about AI and rare birds from the experts on this forum.
Thanks,
I hope this helps!
And
Why did the scarecrow win an award.?
Because he was outstanding in his field.
I like Bard interactions more than the real thing!!
I'm new to the forum, but I'm a big fan of birding and I'm interested in the potential impact of AI on the hobby. I saw the thread about AI-generated images of rare birds, and I wanted to share my thoughts.
I think it's definitely a possibility that AI-generated images could be used to prank or deceive people about rare birds. However, I also think that AI could be used to help birders in a number of ways. For example, AI could be used to develop new identification tools that are more accurate than current methods. AI could also be used to track bird populations and migration patterns, which could help us to better understand and protect these species.
Overall, I think the impact of AI on birding is still uncertain. However, I believe that it has the potential to be both positive and negative. It's important to be aware of the potential risks, but I'm also excited about the potential benefits that AI could bring to the hobby.
I'm looking forward to learning more about AI and rare birds from the experts on this forum.
Thanks,
I hope this helps!
And
Why did the scarecrow win an award.?
Because he was outstanding in his field.
I like Bard interactions more than the real thing!!
dantheman
Bah humbug
Crumbs. Looks like the Peter Jones algorithm is perfect already.
Mysticete
Well-known member
you would need to custom build such an AI right? The stuff I am talking about you can implement now for free and without much technical know how.
Mysticete
Well-known member
Yeah, I didn't say it could do so now, but the capability is growing quickly enough that I could see it being possible within 5 years.
I don't know.
All I know is that the technology makes it difficult to distinguish between human or computer generated data.
Why won't we be able, in a few years time to just tell it to produce checklists in ebird format?
That must be way easier than making photos of birds
jurek
Well-known member
For me, AI images is just another technology. It can be used for fakes, just like Photoshop could be. Just my thoughts, how to cope with AI generated images:
I am too, looking for ways to make AI work, rather than obstruct, birding.
I hope an algorithm could be made which predicts sites and places for birding. it was already done manually by compiling records. Perhaps AI could make it detailed enough to be useful: e.g. find me a walking path within 30 min drive of my home which maximizes possibility of finding a rare bird. Use a compilation of real records, an aerial map of the country, weather forecast and time of the year. A prompt from the user - look especially at bushes, waterbodies and mud flats. I think it is realistic.
But it would require more than a rigid algorithm controlled by Google. A personal software which could be modified and trained or unusual data.
Other way would be to predict how a bird flies. Birds vary in their flight mode - some fly in a straight line, others in a deeper or slower wave, some have long flutter phase, some short, some soar, others cannot soar. I suspect this is controlled simply by birds weight and size and shape of its wings. So an AI algorithm could predict mode of flight of a bird species where it is unknown, e.g. a Kinglet Calypura. Or a long extinct bird like a New Zealand Laughing Owl.
- Be aware of what technology can and cannot do. This also means resisting hype. AI struggles with real wildlife so far.
- Check the outside-the-photo context. If somebody says that he photographed a New Zealand bird in his English garden, this is unlikely.
- Check reliability and interests of the source. An example: about a year ago, Russians released an AI film of Ukrainian president admitting defeat. Ukrainian population simply ignored it. Which says a lot how societies already know to distrust media and resist fakes.
- Don't allow cheats and punish cheats beyond what they might profit from cheating. Societies dealt with cheats this way since forever. They, however, liked stories - fakes which are well understood to be fakes from the start.
- Demand sufficient data that would be too expensive for AI to produce. An example: one photo of a bird can be faked at home, but a 10 sec film not.
- Look for rare but distinct elements. AI is a clever mixing existing sources, so it fails with objects with too few sources. Example: I tried to make AI panorama of Basel city, and failed. It produced a panorama of a generic city. Apparently, there are too few panoramas of Basel city in the internet. This is also the reason why AI birds often fail.
- Look for pieces of image which don't fit together. AI is mixing existing sources, so some pieces will be mixed wrongly. An example: I tried to make AI photo of a tern, and parts of the bird were summer plumage, and other parts winter plumage. Or the AI thrush above has messed-up breast spots.
- Look for logical pieces which aren't there. AI does not understand real world, so will fail. An example is a well known AI mistake of making people with wrong number of fingers, teeth, legs etc.
- People try leaving a room for fake AI in areas which supposedly benefits them. But cheating is unacceptable independent from whom it benefits. An example: some people think that AI Trump riding a dinosaur is OK, but AI Blinken riding a dinosaur would be unacceptable.
- Force big corporations to part with ownership and control of AI. They can make profit from selling it, not renting it. An example is this rumoured Google digital watermark. If it is possible to insert and check a watermark, it will be possible to remove it. Or insert a fake watermark to discredit a real picture. And, since a big part of population distrusts big corporations, a watermark will not be credible to them.
I am too, looking for ways to make AI work, rather than obstruct, birding.
I hope an algorithm could be made which predicts sites and places for birding. it was already done manually by compiling records. Perhaps AI could make it detailed enough to be useful: e.g. find me a walking path within 30 min drive of my home which maximizes possibility of finding a rare bird. Use a compilation of real records, an aerial map of the country, weather forecast and time of the year. A prompt from the user - look especially at bushes, waterbodies and mud flats. I think it is realistic.
But it would require more than a rigid algorithm controlled by Google. A personal software which could be modified and trained or unusual data.
Other way would be to predict how a bird flies. Birds vary in their flight mode - some fly in a straight line, others in a deeper or slower wave, some have long flutter phase, some short, some soar, others cannot soar. I suspect this is controlled simply by birds weight and size and shape of its wings. So an AI algorithm could predict mode of flight of a bird species where it is unknown, e.g. a Kinglet Calypura. Or a long extinct bird like a New Zealand Laughing Owl.
Last edited:
opisska
rabid twitcher
The BirdForum forum has a subforum dedicated to AI and rare birds. In this subforum, birders can discuss the use of artificial intelligence (AI) to identify and track rare birds. Some of the topics that have been discussed in this subforum include:
The AI and rare birds subforum is a valuable resource for birders who are interested in learning more about the use of AI to study and protect rare birds. The subforum is also a good place to share information about rare bird sightings.
- The use of AI-powered apps to identify birds by sight or sound
- The use of AI to track the migration patterns of rare birds
- The use of AI to predict the distribution of rare birds
- The ethical implications of using AI to study rare birds
The AI and rare birds subforum is a valuable resource for birders who are interested in learning more about the use of AI to study and protect rare birds. The subforum is also a good place to share information about rare bird sightings.
opisska
rabid twitcher
(Is anybody here even real anymore? I have actually been seeing this on several forums - new posters chipping in to some random active thread with a very generic "advice" that I am pretty sure was AI generated. But what is the game here? Are they just building trust to eventually post some spam? Or is there a more sinister story?)
dantheman
Bah humbug
It's already pretty hard to keep track - is this you posting, or are you using AI to write the above? (Serious question!)
(ie please can you link to the section you refer to - I couldn't see it at a glance)
I've seen this on BirdForum I believe - randomly generic spurious posts from new users. Don't know if they continued or if they were dealt with by admin ...
There is also a difference between a genuine AI account (like a bot on twitter?) and a real person joining but using AI to generate the content of posts?
Farnboro John
Well-known member
Similar threads
