3 ms·

It's incredible to me how we see NLP models with billions of parameters struggling to act human even on paper, yet somehow many influential people still claim t

It's incredible to me how we see NLP models with billions of parameters struggling to act human even on paper, yet somehow many influential people still claim that the Myers-Briggs type indicator can give predictive value to a 4-bit model of human personality. I get that most people are likely unaware of the measures of complexity involved in these issues, but at the very least I would expect today's intellectual trendsetters to stop giving any further validation to such an obsolete piece of pseudoscience once and for all.

mason55 5y ago
And if you are going to talk about personality traits then at LEAST talk about ones with some scientific backing (the Big 5)
- Noumenon72 5y ago
  The Big 5 replicate, but that's about all; they don't predict, work across cultures, and they're not really orthogonal. https://carcinisation.com/2020/07/04/the-ongoing-accomplishment-of-the-big-five/ https://carcinisation.com/2020/07/04/the-ongoing-accomplishm...
- pseudalopex 5y ago
  4 of the Big 5 traits correlate with the Myers Briggs traits.
ravi-delia 5y ago
Myers-Briggs obviously doesn't represent some fundamental truth about how personalities vary, but I'm willing to bet that no matter how you ask people questions about themselves there will be useful conclusions to draw. Certainly some people are more sociable than others, some more detail-oriented than others, and vary on any other traits you care to name. Given that they are all probably hopelessly correlated with each other, and with whatever variables are actually controlling things underneath, it would be difficult to come up with a test that didn't say something. On a far more speculative note, I bet most of the complex parts of the brain don't vary much between people, or at least don't play into personality much. That's not backed up by anything though, just an idle thought.

mason55 5y ago

And if you are going to talk about personality traits then at LEAST talk about ones with some scientific backing (the Big 5)

Noumenon72 5y ago
The Big 5 replicate, but that's about all; they don't predict, work across cultures, and they're not really orthogonal. https://carcinisation.com/2020/07/04/the-ongoing-accomplishment-of-the-big-five/ https://carcinisation.com/2020/07/04/the-ongoing-accomplishm...
pseudalopex 5y ago
4 of the Big 5 traits correlate with the Myers Briggs traits.

ravi-delia 5y ago

Myers-Briggs obviously doesn't represent some fundamental truth about how personalities vary, but I'm willing to bet that no matter how you ask people questions about themselves there will be useful conclusions to draw. Certainly some people are more sociable than others, some more detail-oriented than others, and vary on any other traits you care to name. Given that they are all probably hopelessly correlated with each other, and with whatever variables are actually controlling things underneath, it would be difficult to come up with a test that didn't say something. On a far more speculative note, I bet most of the complex parts of the brain don't vary much between people, or at least don't play into personality much. That's not backed up by anything though, just an idle thought.