If I were to guess, being pleasantly surprised is just a learned appropriate social response from the expectation of receiving a reward and as such, that social norm is codified sufficiently enough in our writings that it appears in LLMs output.
It’s sort of like all the people who will ask Claude or GPT to validate their complete nonsense and receive unyielding praise for it, the models just learned that this is the best received response based on training data and RL.
I bet these same sorts of expressions can be found in practically every failed attempt as well.
There's historically massive public interest towards discriminating against people on age, gender, or race and many other attributes. Literally happens all the time. Jobs discriminate in hiring, women and men discriminate in dating, apartments discriminate,etc. etc.
So your Human Resources version of reality is actually the opposite of what's true currently and what always has been true.
As a matter of fact there's a great deal of science showing that people feel more comfortable around their own race and age.
Dogs on the other hand do not discriminate and will love you with no bias.
It seems to me that historically there was a high water mark for what people were able to accomplish on their own and all people, dumb to smart, fell somewhere at or below that high water mark.
My opinion is that our current generation of A.I. will just raise the high water mark for what's possible.
Until we have AGI, ASI, or Recursively improving AI....A.I. in it's current form is a tool that adds a constant to some peoples intelligence and an exponent to other peoples intelligence...and many people of all intelligences refuse to use it.
> AI-developed nootropics and germline editing will saturate human intelligence across individuals.
This is a forum, not a Bayesian-updating machine. So common courtesy is for you to take up the burden of proof for your own statements. Why do you think this is the null?
Me too. Too bad I had to abandon it because Mozilla seem more focused on making AI-nonsense than a browser nowadays. No, Firefox, you don’t need AI features. You do need a better UI. You do need smoother scrolling. You do need privacy focus (which automatically means you do not need AI features). You do need to feel like you belong on the platform you’re installed on and not wish everything were Linux. You do need better designers that understand UX is not just “add controls everywhere”.
I remember using lynx at school, since they only let A+ level students use the X11 GUI. And then later on using elinks which was a bit more featureful on my own servers...
So I'm not sure how it knows to be 'surprised' that alone is pretty fascinating.
reply