4 ms·

> The problem with erasing biases is that you cannot look at any statistics. The internet and training set can be free of any form of -ism, and the models would

by datsci_est_2015 8mo ago

> The problem with erasing biases is that you cannot look at any statistics. The internet and training set can be free of any form of -ism, and the models would still be expected to have biases. In fact it's something desirable, because statistical inferences are a valuable tool. I’m sorry but I can’t let you get away with this terrible argument and conclusion. No one argues for completely erasing bias (especially the scientific form of the word bias), that’s a strawman. Strong proponents argue that we should all be aware of our biases, and attempt to adjust our opinions and behavior according to the results of that exercise of self-reflection. Stronger proponents might even argue that the inability to perform this exercise of self-reflection is a path to bigotry. Being racist AF isn’t something that you can excuse with “statistical inference”, and your comment sounds like it’s flirting with that concept. It’s the intellectually juvenile pseudo-philosophy that the techbro scene is absolutely riddled with like a malignant sexually transmitted infection, all the way up to Mu$k and Thi€l. Back to LLM world, the issue is that there is no diversity in its bias: one LLM, one bias. If everyone uses the same dozen or so state-of-the-art LLMs, then all of our processes will have the same dozen or so biases. That would kind of suck if you were a member of a group that those LLMs happened to be biased against. LLMs are also famously not capable of self-reflection, barring the Rube Goldberg machines that people have built on top of them to simulate thought processes.

WarmWash 8mo ago
>The weak link is humans not recognizing that that doesn't mean there cannot be good Italian food in south Texas. A logical hurdle I don't see AI having any problem with. Like your argument mentions, the problem is with human brains, not AI. AI is already plainly miles ahead of most humans in understanding nuance. What will be inescapable though, is trying to be an Italian restaurant that can compete for customers in a south Texas environment will just intrinsically be much more difficult than being a Mexican place. Even the most honest morally pure AI will tell people "When in south texas, you gotta have their mexican food"
- datsci_est_2015 8mo ago
  > AI is already plainly miles ahead of most humans in understanding nuance. That’s a fiery hot take, unless the words “understanding” and “nuance” are doing some concerningly heavy lifting. Either that or you have an incredibly low opinion of “most humans” that borders on misanthropy. > What will be inescapable though, is trying to be an Italian restaurant that can compete for customers in a south Texas environment will just intrinsically be much more difficult than being a Mexican place. Even the most honest morally pure AI will tell people "When in south texas, you gotta have their mexican food" This line of argumentation is bizarre that I can only imagine it was chosen by the OP because it sounded more innocent than something like “AI putting black men in jail because it was trained on 4chan”. Also what is “moral purity”? Sounds condescending to the concept of fighting unjust bias.
  - WarmWash 8mo ago
    [flagged]
hattmall 8mo ago
Ok, but what if you're a non-emotional system, whose biases are generated from objective statistical data? Then you become aware of those biases, and "adjust our opinions and behavior". You are just introducing inefficiency, and it speaks to the OPs point of: > The problem with erasing biases is that you cannot look at any statistics. If you can't use the statistics to generate biases then what is the purpose of building an inefficient processor. Not only is it inefficient because of ignoring the statistical data, it's inefficiency is compounded by the fact that you have to go out of your way to add extra layers in order to mitigate the observable statistical inference.
- datsci_est_2015 8mo ago
  > Ok, but what if you're a non-emotional system, whose biases are generated from objective statistical data? Objective statistical data doesn’t exist, that’s Data Science / Statistics 101. Your sample always has a bias, unless your sample is: everything, always, how it’s been, and how it always will be. I don’t really know what inefficiency has to do with anything, wish I could respond to the rest of your comment.

WarmWash 8mo ago

>The weak link is humans not recognizing that that doesn't mean there cannot be good Italian food in south Texas. A logical hurdle I don't see AI having any problem with. Like your argument mentions, the problem is with human brains, not AI. AI is already plainly miles ahead of most humans in understanding nuance. What will be inescapable though, is trying to be an Italian restaurant that can compete for customers in a south Texas environment will just intrinsically be much more difficult than being a Mexican place. Even the most honest morally pure AI will tell people "When in south texas, you gotta have their mexican food"

datsci_est_2015 8mo ago
> AI is already plainly miles ahead of most humans in understanding nuance. That’s a fiery hot take, unless the words “understanding” and “nuance” are doing some concerningly heavy lifting. Either that or you have an incredibly low opinion of “most humans” that borders on misanthropy. > What will be inescapable though, is trying to be an Italian restaurant that can compete for customers in a south Texas environment will just intrinsically be much more difficult than being a Mexican place. Even the most honest morally pure AI will tell people "When in south texas, you gotta have their mexican food" This line of argumentation is bizarre that I can only imagine it was chosen by the OP because it sounded more innocent than something like “AI putting black men in jail because it was trained on 4chan”. Also what is “moral purity”? Sounds condescending to the concept of fighting unjust bias.
- WarmWash 8mo ago
  [flagged]

hattmall 8mo ago

Ok, but what if you're a non-emotional system, whose biases are generated from objective statistical data? Then you become aware of those biases, and "adjust our opinions and behavior". You are just introducing inefficiency, and it speaks to the OPs point of: > The problem with erasing biases is that you cannot look at any statistics. If you can't use the statistics to generate biases then what is the purpose of building an inefficient processor. Not only is it inefficient because of ignoring the statistical data, it's inefficiency is compounded by the fact that you have to go out of your way to add extra layers in order to mitigate the observable statistical inference.

datsci_est_2015 8mo ago
> Ok, but what if you're a non-emotional system, whose biases are generated from objective statistical data? Objective statistical data doesn’t exist, that’s Data Science / Statistics 101. Your sample always has a bias, unless your sample is: everything, always, how it’s been, and how it always will be. I don’t really know what inefficiency has to do with anything, wish I could respond to the rest of your comment.