So they’re taking “AI” predictions and using them as measurements taken from the underlying process?
I’ve been watching this “AI” thing and it seems, in some circles, to have begun to be elevated, cult-like, to the level of some kind of mysterious deity.
Would not be surprised if the AI engines are fed "data", which is output from previous models to create "predictions" which is then fed into the models to create more data. It's the circle of life.
I saw a clip not long ago of that disgruntled old nazi Klaus Schwab chortling about how the predictive powers of AI would soon make elections obsolete because AI would simply be able to tell us who would get elected and therefore we wouldn't even need to have an election. What an asshole.
Is this pattern of predictive modeling with reshuffled or AI-generated data a result of laziness or lack of funding? Too lazy to get more real data points? Or not enough funding for the field work involved in getting more real data points?
Without your easy to understand explanations, I admit I’d most likely be bamboozled by the “experts” and their p- values. Reading your posts helps me understand how the “expert class” has caused so much havoc in our world. Thanks for helping us see behind the curtain. The Great Oz is not so great.
'Their idea is simple. Use “machine learning” to make predictive models conditioned on data, and use the predictions as if it were “new” data to produce shinier P-values.'
Although the bootstrap is sometimes sold as a way to get uncertainty measures in cases with scarce data, I would argue that it is not the bootstrap main purpose. The bootstrap is a computational way of getting uncertainty measures where there is no good analytical method. By the way, I like the idea of predictive inference, which is actually in the spirit of machine learning rather than parametric modelling.
So they’re taking “AI” predictions and using them as measurements taken from the underlying process?
I’ve been watching this “AI” thing and it seems, in some circles, to have begun to be elevated, cult-like, to the level of some kind of mysterious deity.
There’s plenty of danger lurking in “AI” worship.
The jump from Artificial to Almighty is not particularly cavernous for the the Tik-Tok culture
What is the AI using for its predictions? Models? Designed by humans?
Would not be surprised if the AI engines are fed "data", which is output from previous models to create "predictions" which is then fed into the models to create more data. It's the circle of life.
I saw a clip not long ago of that disgruntled old nazi Klaus Schwab chortling about how the predictive powers of AI would soon make elections obsolete because AI would simply be able to tell us who would get elected and therefore we wouldn't even need to have an election. What an asshole.
Reminded me of Zeigler and Evan’s “In the year 2525”
Is this pattern of predictive modeling with reshuffled or AI-generated data a result of laziness or lack of funding? Too lazy to get more real data points? Or not enough funding for the field work involved in getting more real data points?
Without your easy to understand explanations, I admit I’d most likely be bamboozled by the “experts” and their p- values. Reading your posts helps me understand how the “expert class” has caused so much havoc in our world. Thanks for helping us see behind the curtain. The Great Oz is not so great.
Seems like a natural extension to the climatologists claiming a new supercomputer simulation is running an experiment.
'Their idea is simple. Use “machine learning” to make predictive models conditioned on data, and use the predictions as if it were “new” data to produce shinier P-values.'
When confronted with situations like this, my 1985 Mac 512 K would throw up the complaint "Too Much Recursion." The graphic in this Note is from a later time, but gets the idea across: https://substack.com/profile/4958635-tardigrade/note/c-46233018
Efron beckons -)
Although the bootstrap is sometimes sold as a way to get uncertainty measures in cases with scarce data, I would argue that it is not the bootstrap main purpose. The bootstrap is a computational way of getting uncertainty measures where there is no good analytical method. By the way, I like the idea of predictive inference, which is actually in the spirit of machine learning rather than parametric modelling.