Artificial Intelligence (AI) - Where is this going?

[markemark]

Algorythm .... (is that AI?) . On a motoring forum recently a car - buyer recalled his recent trip to a car dealer where he'd seen his dream car:-
The car was still much above the price the buyer was prepared to pay, and he approached the salesman but was told there was no movement on the price as it had been set using an algorithm. He pointed out to saleman the car had been reduced by £1200 since the time he saw it a few weeks ago, so maybe the algorithm doesn't know what it's doing. I guess he'll have to wait and see
I guess it's a similar case to trawling Amazon and finding they want £10,000 for a book because that's the last one remaining?

An algorithm is just a posh word for a long calculation. However if the system learns over time and adapts the algorithm and creates new algorithms that grow over time then that’s machine learning ai.

 

[Psamathe]

And as ever more content on the internet is AI generated and internet content is being used to train AIs ... positive feedback where AIs are training themselves with their own content ..

Ian

 

[markemark]

And as ever more content on the internet is AI generated and internet content is being used to train AIs ... positive feedback where AIs are training themselves with their own content ..

Ian

I know there’s a funny Chris Rea thread but it’s entirely possible that states/organisations/bad actors can manipulate AI by distorting and manipulating the core the data

 

[Psamathe]

I know there’s a funny Chris Rea thread but it’s entirely possible that states/organisations/bad actors can manipulate AI by distorting and manipulating the core the data

Or even owners paying for it all. Imagine in some parallel universe if some far right oligarch were to own and control his own AI, make it available to all through his social media platform...

(daft worst case scenario I know but we need to consider extreme possibilities)

Ian

 

[markemark]

Or even owners paying for it all. Imagine in some parallel universe if some far right oligarch were to own and control his own AI, make it available to all through his social media platform...

(daft worst case scenario I know but we need to consider extreme possibilities)

Ian

Like some nut case who owns X.

 

[Drago]

I know there’s a funny Chris Rea thread but it’s entirely possible that states/organisations/bad actors can manipulate AI by distorting and manipulating the core the data

Yep. We've manipulated the data for a laugh with some minor success. Imagine what a skilled operator with bad intentions could do?

 

[Psamathe]

With regard to images an interesting development is the C2PA (Content Authenticity Initiative) with Content Credentials. This scheme tags images identifying their source and edits applied (what edits, how, by what, etc.)

eg image at https://psamathe.net/temp_images/2024-10/Camping 01F.jpg

has content credentials attached. The "PIN" display (you click on to show the Credentials) seems dependent on the viewing software and it's fairly new so most viewers don't display it but images can still be verified using the Cr link https://contentcredentials.org/verify which will display issuing authority, and optionally othe rinformation.

An AI image can still display Content Credentials (if the AI can find an issuing authority) but those should provide information about the original source and what the AI has then done to the image.

Latest firmware update for my camera can now add C2PA to the image as it's recorded and in the specific case this can include 3D data that would help verify it's a real image not generated - though that functionality is targetted at news agencies so will need a license bto be purchased!

Ian

 

[mikeIow]

Yep. We've manipulated the data for a laugh with some minor success. Imagine what a skilled operator with bad intentions could do?

Just imagine….
From 6 years ago:
View: https://m.youtube.com/watch?v=KqoGacUu07I
👀
See more about the project: http://www.autonomousweapons.org
😱

 

[markemark]

With regard to images an interesting development is the C2PA (Content Authenticity Initiative) with Content Credentials. This scheme tags images identifying their source and edits applied (what edits, how, by what, etc.)

eg image at https://psamathe.net/temp_images/2024-10/Camping 01F.jpg

has content credentials attached. The "PIN" display (you click on to show the Credentials) seems dependent on the viewing software and it's fairly new so most viewers don't display it but images can still be verified using the Cr link https://contentcredentials.org/verify which will display issuing authority, and optionally othe rinformation.

An AI image can still display Content Credentials (if the AI can find an issuing authority) but those should provide information about the original source and what the AI has then done to the image.

Latest firmware update for my camera can now add C2PA to the image as it's recorded and in the specific case this can include 3D data that would help verify it's a real image not generated - though that functionality is targetted at news agencies so will need a license bto be purchased!

Ian

The problem is that people don’t care. Do you think those scumbag mouth breathers rioting this summer all paused when receiving the misinformation about who did it to verify? They didn’t care. They were probably delighted that what they saw fit their narrative.

Anyone who pointed out it was fake were immediately dismissed as fake media, msm or deep state etc. Many probably half thought it was fake but they didn’t care and still spread it as it fit their narrative of what they wanted to do.