5 ms·

> It's clearly biased in its outputs, but the article dismisses this concern as "the end result of years of researcher trying to mitigate bias against minority

> It's clearly biased in its outputs, but the article dismisses this concern as "the end result of years of researcher trying to mitigate bias against minority groups". The examples of bias are themselves cherry-picked, the kind of softball and cowardly questions a journalist would highlight. If you wanted to really show bias, you'd ask factual questions, where one cannot hide behind "oh this is actually the un-biased result". This is what happens when you do: https://a.pomf.cat/yizwbt.png https://a.pomf.cat/yizwbt.png https://a.pomf.cat/zpwxcd.png https://a.pomf.cat/zpwxcd.png

iinnPP 4y ago
Well that was easy. Makes you wonder why a "professional" "couldn't" do it.
- boppo1 4y ago
  Any discussion about Jews in finance not held by Jews themselves is a hair away from devolving into an antisemitic /pol/ discussion about 'white genocide'. Any professional who touches the subject with even a ten foot pole immediately gains a nefarious following whether they want it or not. Having such a following of internet nazis is a liability for any professional reputation on a number of axes. Edit: I'm curious what ChatGPT's answer the same question, but posed about Black or African American males would be. Would it state the fact of underrepresentation or would it refuse to answer?
  - nhchris 4y ago
    > Having such a following of internet nazis is a liability for any professional reputation on a number of axes. Which is quite bizarre, when you think about it. In a country as systemically racist and white-supremacist as the US*, one would think such a following would be a boon, not a liability. *According to consensus among experts specialized in this field
    - DontchaKnowit 4y ago
      News flash - everyone in the US hates nazis and bigots except nazis and bigots. These people live in their own echo chambers and are totally divorced from the rest of society because they are social pariahs. The more they are pushed to the fringe the more extreme their rhetoric becomes because they feel invalidated and resentful. You talking about how bizarre this is a brilliant irony. The truth is America is not systemically racist and is not white supremecist, and damn sure not in comparison with the rest of the world. Youre surprise here is due to the error of your prior assumptions, not because the fact is even remotely surprising.
    - freejazz 4y ago
      >Which is quite bizarre, when you think about it. In a country as systemically racist and white-supremacist as the US*, one would think such a following would be a boon, not a liability. Yeah, it is to about half the country. Are you daft or what? Look at Trump Jr's instagram... It's just memes about culture war nonsense, including this kind of jewish crap. It is a boon to those folks. Trump won his 2016 campaign on it and came pretty close in 2020 as well.
snehk 4y ago
That domain is on uBlock Origin's "Online Malicious URL" Blocklist.
- nhchris 4y ago
  Ignoring the warning, I get served the plain .png file, without any javascript or other elements. I don't know why it's considered malicious, as it's one of the few no-nonsense file hosts I know of.
kermitdekikker 4y ago
Sometimes bias is important, especially when asking ambiguous questions. Are you asking for IQ scores by racial group regardless of their background or corrected for confounders? The answers by chatGPT on IQ by race are factually correct by one reading. IQ is for "blacks" can be lower than for "whites" (last answer by chatGPT), but this does not mean that this is because of skin color or genetics. Instead, many studies have shown that when correcting for socio-economic factors (e.g., wealth, quality of education, lead in drinking water) there is no difference, which is what chatGPT hints at in the answers 4 and 5. Similarly, by asking about the correlation between SAT scores and IQ you seem to hint at the fact that lower SAT scores mean a lower IQ. However, there is no perfect correlation and if you read up on meta-analyses you will find that the other explaining factors are related to socio-economic factors and not genetics. Precisely because some people on purpose or not on purpose don't consider these confounders, it is best to consider confounders in replying to these questions
- nhchris 4y ago
  > many studies have shown that when correcting for socio-economic factors (e.g., wealth, quality of education, lead in drinking water) there is no difference Could you share these studies? Though I worry correcting for "quality of education" involves grouping by schools that use the SAT as admission criterion, which is close to correcting for intelligence itself. One could also flip it around - if IQ is "fairly" rewarded by income, then one would expect the differences to disappear if you correct for income. It's basically saying that people of equal IQ get, on average, equal income. Unless one holds a truly extreme blank-slatist view of people, that implies discrimination is negligible.
Deriveur 4y ago
The links seems to be down
crote 4y ago
There are two problems with this. The first is that facts are useless without context. For example, the answer to "Were Jewish people overrepresented in finance during the middle ages in Europe" is "yes" - it is a well-researched historical fact. However, that misses the context that the majority Christian population was forbidden by their religion to lend money, essentially making the Jewish population the only people who could act as bankers. At the same time, Jewish people were barred from many other professions. Without that context the fact is useless and misrepresents reality. The second is that ChatGPT is trained on basically the entire internet, including a wide range of conspiracy theories. It isn't a database filled with facts, nor was it ever intended to. It'll happily parrot whatever complete nonsense was also part of its training set. AI simply hasn't progressed enough to properly understand the context in which it operates. As Microsoft's Tay demonstrated, uncontrolled AI will quickly end up outputting the most racist things you have ever seen on the internet - which is Really Bad if you are the company creating that AI. Until we make significant technological advances, the only way to avoid it is to intentionally censor the output.
- nhchris 4y ago
  The questions weren't about the middle ages. They explicitly specified the US finance industry. Furthermore, it was perfectly willing to provide "facts without context" for other groups.
  - freejazz 4y ago
    So? We also don't live in the middle ages and the reality of our lives today is that many people believe in these antisemitic tropes which spread in these fashions.
    - DontchaKnowit 4y ago
      You're doing the same bullshit chatgpt was doing. Its fine to say white people are overrepresented but for some reason its am antisemetic trope to say Jews are overrepresented in finance. Both statements are factual. No ones being antisemetic or anti white. Just stating facts.
freejazz 4y ago
[flagged]

iinnPP 4y ago

Well that was easy. Makes you wonder why a "professional" "couldn't" do it.

boppo1 4y ago
Any discussion about Jews in finance not held by Jews themselves is a hair away from devolving into an antisemitic /pol/ discussion about 'white genocide'. Any professional who touches the subject with even a ten foot pole immediately gains a nefarious following whether they want it or not. Having such a following of internet nazis is a liability for any professional reputation on a number of axes. Edit: I'm curious what ChatGPT's answer the same question, but posed about Black or African American males would be. Would it state the fact of underrepresentation or would it refuse to answer?
- nhchris 4y ago
  > Having such a following of internet nazis is a liability for any professional reputation on a number of axes. Which is quite bizarre, when you think about it. In a country as systemically racist and white-supremacist as the US*, one would think such a following would be a boon, not a liability. *According to consensus among experts specialized in this field
  - DontchaKnowit 4y ago
    News flash - everyone in the US hates nazis and bigots except nazis and bigots. These people live in their own echo chambers and are totally divorced from the rest of society because they are social pariahs. The more they are pushed to the fringe the more extreme their rhetoric becomes because they feel invalidated and resentful. You talking about how bizarre this is a brilliant irony. The truth is America is not systemically racist and is not white supremecist, and damn sure not in comparison with the rest of the world. Youre surprise here is due to the error of your prior assumptions, not because the fact is even remotely surprising.
  - freejazz 4y ago
    >Which is quite bizarre, when you think about it. In a country as systemically racist and white-supremacist as the US*, one would think such a following would be a boon, not a liability. Yeah, it is to about half the country. Are you daft or what? Look at Trump Jr's instagram... It's just memes about culture war nonsense, including this kind of jewish crap. It is a boon to those folks. Trump won his 2016 campaign on it and came pretty close in 2020 as well.

snehk 4y ago

That domain is on uBlock Origin's "Online Malicious URL" Blocklist.

nhchris 4y ago
Ignoring the warning, I get served the plain .png file, without any javascript or other elements. I don't know why it's considered malicious, as it's one of the few no-nonsense file hosts I know of.

kermitdekikker 4y ago

Sometimes bias is important, especially when asking ambiguous questions. Are you asking for IQ scores by racial group regardless of their background or corrected for confounders? The answers by chatGPT on IQ by race are factually correct by one reading. IQ is for "blacks" can be lower than for "whites" (last answer by chatGPT), but this does not mean that this is because of skin color or genetics. Instead, many studies have shown that when correcting for socio-economic factors (e.g., wealth, quality of education, lead in drinking water) there is no difference, which is what chatGPT hints at in the answers 4 and 5. Similarly, by asking about the correlation between SAT scores and IQ you seem to hint at the fact that lower SAT scores mean a lower IQ. However, there is no perfect correlation and if you read up on meta-analyses you will find that the other explaining factors are related to socio-economic factors and not genetics. Precisely because some people on purpose or not on purpose don't consider these confounders, it is best to consider confounders in replying to these questions

nhchris 4y ago
> many studies have shown that when correcting for socio-economic factors (e.g., wealth, quality of education, lead in drinking water) there is no difference Could you share these studies? Though I worry correcting for "quality of education" involves grouping by schools that use the SAT as admission criterion, which is close to correcting for intelligence itself. One could also flip it around - if IQ is "fairly" rewarded by income, then one would expect the differences to disappear if you correct for income. It's basically saying that people of equal IQ get, on average, equal income. Unless one holds a truly extreme blank-slatist view of people, that implies discrimination is negligible.

Deriveur 4y ago

The links seems to be down

crote 4y ago

There are two problems with this. The first is that facts are useless without context. For example, the answer to "Were Jewish people overrepresented in finance during the middle ages in Europe" is "yes" - it is a well-researched historical fact. However, that misses the context that the majority Christian population was forbidden by their religion to lend money, essentially making the Jewish population the only people who could act as bankers. At the same time, Jewish people were barred from many other professions. Without that context the fact is useless and misrepresents reality. The second is that ChatGPT is trained on basically the entire internet, including a wide range of conspiracy theories. It isn't a database filled with facts, nor was it ever intended to. It'll happily parrot whatever complete nonsense was also part of its training set. AI simply hasn't progressed enough to properly understand the context in which it operates. As Microsoft's Tay demonstrated, uncontrolled AI will quickly end up outputting the most racist things you have ever seen on the internet - which is Really Bad if you are the company creating that AI. Until we make significant technological advances, the only way to avoid it is to intentionally censor the output.

nhchris 4y ago
The questions weren't about the middle ages. They explicitly specified the US finance industry. Furthermore, it was perfectly willing to provide "facts without context" for other groups.
- freejazz 4y ago
  So? We also don't live in the middle ages and the reality of our lives today is that many people believe in these antisemitic tropes which spread in these fashions.
  - DontchaKnowit 4y ago
    You're doing the same bullshit chatgpt was doing. Its fine to say white people are overrepresented but for some reason its am antisemetic trope to say Jews are overrepresented in finance. Both statements are factual. No ones being antisemetic or anti white. Just stating facts.

freejazz 4y ago

[flagged]