The books "The Age of Surveillance Capitalism", Christopher Wylie's "Mindfuck", and reading about nudges [hide=*](and henceforth becoming irrationally angry when you see those fucking flies painted on urinals if you're a bloke...)[/hide] might help you get a broader idea of how they work societally.
The specifics of the algorithms are notoriously "black box", so over and above suggestions (like in Mindfuck) that people try to estimate Big 5 personality traits from posts, and perform automatic advertising controlled experiments on their platforms (eg Amazon and Facebook) to maximise click through and conversion rates... I don't think that can be said. Exactly what the AI thinks is hard to tell, even if you're the analyst that's fit the model.
Unless Google etc. are sitting on unprecedented "interpretable AI" advances, it'll be mostly blackboxes tuned by controlled experiments on social media, Amazon, credit ratings stuff... Depends who's sharing what with who really, but we have no idea eh?
Edit: you'll probably see that the above isn't really talking about minds, it's talking about observables that human agents generate. How well those observables can be combined to model the agent depends upon the observables, the model, and the agent. And even then, the model's to predict stuff like click through rates and what ads people should be shown, not like... "how does this person react to stress?" or other character traits' manifestations in context.
Comments (3)
The books "The Age of Surveillance Capitalism", Christopher Wylie's "Mindfuck", and reading about nudges [hide=*](and henceforth becoming irrationally angry when you see those fucking flies painted on urinals if you're a bloke...)[/hide] might help you get a broader idea of how they work societally.
The specifics of the algorithms are notoriously "black box", so over and above suggestions (like in Mindfuck) that people try to estimate Big 5 personality traits from posts, and perform automatic advertising controlled experiments on their platforms (eg Amazon and Facebook) to maximise click through and conversion rates... I don't think that can be said. Exactly what the AI thinks is hard to tell, even if you're the analyst that's fit the model.
Unless Google etc. are sitting on unprecedented "interpretable AI" advances, it'll be mostly blackboxes tuned by controlled experiments on social media, Amazon, credit ratings stuff... Depends who's sharing what with who really, but we have no idea eh?
Edit: you'll probably see that the above isn't really talking about minds, it's talking about observables that human agents generate. How well those observables can be combined to model the agent depends upon the observables, the model, and the agent. And even then, the model's to predict stuff like click through rates and what ads people should be shown, not like... "how does this person react to stress?" or other character traits' manifestations in context.
Web cookies? They can be used to analyze your clickstream, which gives clues to your mind.
then that program will eventually be able to do anything a human can do
copying is more important then inventing