AI Fails (9 Viewers)

Ad: This forum contains affiliate links to products on Amazon and eBay. More information in Terms and rules

Lucky13 said:
View attachment 889448
Click to expand...
This is a "science fiction" channel methinks. As such I give the illustration a pass. The story, however, is probably AI slop and narrated by an AI voice. There's a couple of channels that have AI narration but the stories are written by actual people. One is written and narrated by humans though the illustrations are AI.
 
SaparotRob said:
This is a "science fiction" channel methinks. As such I give the illustration a pass. The story, however, is probably AI slop and narrated by an AI voice. There's a couple of channels that have AI narration but the stories are written by actual people. One is written and narrated by humans though the illustrations are AI.
Click to expand...

I've noticed that whenever I listen some of the stories, always the same names, the same professions, etc., etc... 😆😂
There are some half decent though....
 
Capture+_2026-08-05-18-44-30.png
 
fubar57 said:
loire nieuport ln.401
Click to expand...
That seems to be how the tail hook was installed on the Naval versions of the A/C. Remember there were land based versions of the A/C that would not have had the tail hook. But I do not claim to be an expert on the type. According to the source page only 15 of the 401 versions were built with folding wings and naval equipment, 45 examples of the land based 411 models were built, so going by the numbers more pictures of the 411 models are probably available that would not have the arrester hooks mounted.
1786028177134.png

Loire-Nieuport LN.401/LN.411 (LN.40) Carrierborne Dive Bomber Aircraft

Page details technical specifications, development, and operational history of the Loire-Nieuport LN.401/LN.411 (LN.40) Carrierborne Dive Bomber Aircraft including pictures.
www.militaryfactory.com
 
Last edited:
Sabrejet said:
Every time someone (in a bored moment) decides to use AI, another data centre is justified. We need to understand the impacts of this plague.
Click to expand...
I totally agree on this. However, I am realistic and AI is staying here for good. After first being a pioneer in Deep Learning (basis of the current LLM's) in my company, I then started to fight the AI hype, and now I can only join and hope I will not drown. I you can't beat them, join them I am afraid. Doesn't make me less critical of the stuff though.
OldGeezer said:
I tried to use AI for Good instead of Evil yesterday, by taking one of the XP-67 photos of the airplane, pointing at the camera, with engines running and telling the tool to generate a video with the propellers actually spinning. I was careful to say that the one on the left ("of the photo" I assumed it would understand, so I didn't specifically state that) should rotate counterclockwise and the one on the right (again, assuming it would understand that was the photo's right) rotating clockwise. On XP-67 that's kind of important since wind tunnel testing had shown that this was the only way the airplane would stay adequately stable in landing mode or something. So what did the AI tool do? The exact opposite of what I told it. Oh well, I thought, maybe it's so smart that it knows what an airplane is and that "right" and "left" mean the airplane's right and left, not the photo's. So I reversed the instructions, and what did I get? No change at all. I kept tweaking the instructions without improving things, in fact I did at one point have both propellers rotating in the same direction but that was no help either. My final attempt to tell it how to change its product made it cut one blade off of the RH engine, so the LH one had the correct 4 but the RH engine was swinging just 3. And so, 1600 "credits" poorer, I finally gave up. It's too bad, too, because apart from the backwards prop rotations, the clip was pretty good. My next thought was to try putting it into a video editor and just reverse the clip, which would fix the rotation direction problems. But since the AI tool decided to also have the airplane taxi toward the camera, which I did NOT tell it to do, running the clip backward would have the props somehow pushing the airplane backward...
Click to expand...
So, for the ones who do want to use AI, maybe some explanation. AI seems intelligent (after all, that's what the I stands for), but it's actually not. It's a very sophisticated calculation model for pattern recognition and creating. Everything falls and stands with the context size. Think of it like a person with dementia. If your sentences are short enough, they will remember the start and end of the question and understand what you are asking. But if you ask an person with dementia an question that is longer then their memory span, they will not fully comprehend. AI is similar, although their memory span is usually a bit larger than that of a person with dementia. The "memory span" of an AI model is called the "maximum context length".
Some models can work with a larger context then others, that is basically the difference between all of them, apart from the training. If you make the context too big, it will "get" only part of it. What part is anybody's guess and can lead to surprising results, hence the "hallucination" part. The free models invariably have smaller max context than the more expensive ones.

So what does this mean to anybody wanting to use an AI model? Feed the model a question at a time and limit the "context" you give it. So break your question in parts and feed it part at the time. Make sure you stay within the context limits. Every part of the "conversation" you have with the LLM will add to the used context. At one point, the model starts to make mistakes again that it had made earlier and starts to forget things. This is a clear sign you have exceeded the context limits of the model. The only thing you can do then is save the end result at that point and start a new chat with the bot, so you start with a new clean context. Keep your questions small and well scoped. It's the only way to get a decent answer from AI. If you nurse it like that, always remember the limits of a person with dementia, than it can do some useful stuff. If it's efficient or not? That totally depends on the problem you give it and the situation you're in.

At my work, some people only read the AI created summary of scientific articles, claiming it saves them reading time. In the mean time could have read the introduction and conclusion as quickly as that, without spending water and electricity that could have been better used. Also they wouldn't have had the filtered and biased interpretation that a hallucinating model might have had while creating the summary and made up their minds themselves.
 
You must log in or register to reply here.

Users who are viewing this thread

  • ... and 2 more.
    • Total: 10 (members: 3, guests: 7)

    Similar threads

    AMC
    AI here now!!….
    Replies
    0
    Views
    269
    AMC
    AMC
    O
    Finally found a use for AI...maybe?
    Replies
    1
    Views
    372
    Sabrejet
    Sabrejet
    CHen10
    Can you use AI prompts to make aviation ww2 pictures?
    Replies
    14
    Views
    3K
    GrauGeist
    GrauGeist
    Wurger
    Airframes' Birthday !
    Replies
    17
    Views
    2K
    Airframes
    Airframes
    rob23
    Airshow question
    Replies
    2
    Views
    715
    rob23
    rob23
    Back