Mr. A., thank you for your research and comilation. It sent me to the MIT AI/CS lab and, in particular, a paper on the site about AI being confounded by "overinterpretation." Turns out machines can and do make perfectly "unbiased" predictions that have zero confidence attached. THIS IS THE RULE not the exception when serious stuff is …
Mr. A., thank you for your research and comilation. It sent me to the MIT AI/CS lab and, in particular, a paper on the site about AI being confounded by "overinterpretation." Turns out machines can and do make perfectly "unbiased" predictions that have zero confidence attached. THIS IS THE RULE not the exception when serious stuff is happening and decisions MUST be made in a very short period of time. Not exactly suitable for little things like driverless cars, let alone governing self-reproducing economic system. Bottom line is that humans with access to restricted/mixed stochastic regression techniques, full information maximum likelihood methods and spectral analysis can do BETTER than machines using these very same tools of statistical inference. Hoo wah !!!
Mr. A., thank you for your research and comilation. It sent me to the MIT AI/CS lab and, in particular, a paper on the site about AI being confounded by "overinterpretation." Turns out machines can and do make perfectly "unbiased" predictions that have zero confidence attached. THIS IS THE RULE not the exception when serious stuff is happening and decisions MUST be made in a very short period of time. Not exactly suitable for little things like driverless cars, let alone governing self-reproducing economic system. Bottom line is that humans with access to restricted/mixed stochastic regression techniques, full information maximum likelihood methods and spectral analysis can do BETTER than machines using these very same tools of statistical inference. Hoo wah !!!
Here is the link to the MIT paper: https://www.csail.mit.edu/news/nonsense-can-make-sense-machine-learning-models.
BOOM!