There are hundreds of obvious question pairs that can be used to evaluate cognitive biases in an AI - race, religion, age, gender, education/major, profession, institutional affiliation, nationality, species ...
Of course, that only works until the AI can understand the question: "Did you hear what you just said?"
Perhaps more telling, you do realize that the ChatGBT LECTURED you?
I found a research who tested 6,700 negative statements against 75 subgroups to see if there were variations in how the system defined that statement as "hateful."
I'll report this within a week or so, but the short answer is that they treat different types of people and, yes, political ideologies very different.
I see the questions submitted to the ChatGBT AI and most of the conclusions reached are identical to those submitted and reached by writer Glenn Greenwald in his RumbleCAST of February 7, 2023.
With both of you, as independent journalists warning Americans of the potential dangers of group think in the form of AI influencing public discussion and discourse you both should consider a collaborative effort.
It is going to be very challenging to build alternatives. AI is so expensive to build and operate. I hope someone is successful in this endeavor.
The public has a misconception that AI will be unbiased as it is a logical thinking machine. I think it is important to bring awareness that this is in fact not the reality.
FYI, I recently wrote about AI Bias from the stand point of proving the impossibility that AI can be non-biased.
A good article: Christians Must Enter the AI Arms Race, By: Andrew Torba
https://news.gab.com/2023/01/christians-must-enter-the-ai-arms-race/
There are hundreds of obvious question pairs that can be used to evaluate cognitive biases in an AI - race, religion, age, gender, education/major, profession, institutional affiliation, nationality, species ...
Of course, that only works until the AI can understand the question: "Did you hear what you just said?"
Perhaps more telling, you do realize that the ChatGBT LECTURED you?
I found a research who tested 6,700 negative statements against 75 subgroups to see if there were variations in how the system defined that statement as "hateful."
I'll report this within a week or so, but the short answer is that they treat different types of people and, yes, political ideologies very different.
I see the questions submitted to the ChatGBT AI and most of the conclusions reached are identical to those submitted and reached by writer Glenn Greenwald in his RumbleCAST of February 7, 2023.
With both of you, as independent journalists warning Americans of the potential dangers of group think in the form of AI influencing public discussion and discourse you both should consider a collaborative effort.
SOURCE: https://rumble.com/v28oi66-system-update-36.html
It is going to be very challenging to build alternatives. AI is so expensive to build and operate. I hope someone is successful in this endeavor.
The public has a misconception that AI will be unbiased as it is a logical thinking machine. I think it is important to bring awareness that this is in fact not the reality.
FYI, I recently wrote about AI Bias from the stand point of proving the impossibility that AI can be non-biased.
https://dakara.substack.com/p/ai-the-bias-paradox
Google Panics Over ChatGPT: https://odysee.com/@Coldfusion:f/google-panics-over-chatgpt-the-ai-wars:a
ChatGPT is NOT HAL or Starfleet computers from Star Trek: Not Indie minded
NOT a true Dbase.
If you ask OPEN AI this question you may be shocked by its answer.
“Give me 40 examples of communities or groups OPEN AI is trying to protect?”
https://greenwald.locals.com/upost/3477339/chatgpt-is-straight-up-malware