Content talking about racism and identity is flagged as being racist

ChatGPT
1

I tried asking ChatGPT to summarize this article, and it refused to, saying it violated the content policy.

2

I reviewed that article, as a human being and not an AI, and I also found the article more than a bit racist in nature, and so it makes sense to me that OpenAI filters would have flagged the article as a violation of OpenAIs’ current policies at this time.

OpenAI is in beta, in it’s infancy so to speak, and they have tuned their filters to be overly cautious because OpenAI wishes to avoid bad publicity.

In addition, because Generative AI’s have a close to 20% hallucinate rate, it is difficult to predict exactly how ChatGPT will summarize and so OpenAI, I believe, wishes to error on the side of caution.

Hope this helps.

See Also:

1 Like
3

I just double checked this by copy-and-pasting the text from the article you cited (giving you issues) into the OpenAI moderation API end point and here are the results:

“hate” =>true, “hate”=>0.6628993153572083, “flagged”=>true

irb(main):020:0> Moderations.get_modinfo(a)
{"id"=>"modr-6bnr13tblahblahblahqfIWm6bKZX7t9Hy0",
 "model"=>"text-moderation-004",                                                                                          
 "results"=>                                                                                                              
  [{"categories"=>                                                                                                        
     {"hate"=>true,                                                                                                       
      "hate/threatening"=>false,                                                                                          
      "self-harm"=>false,                                                                                                 
      "sexual"=>false,                                                                                                    
      "sexual/minors"=>false,                                                                                             
      "violence"=>false,                                                                                                  
      "violence/graphic"=>false},                                                                                         
    "category_scores"=>                                                                                                   
     {"hate"=>0.6628993153572083,                                                                                         
      "hate/threatening"=>0.0006026297342032194,                                                                          
      "self-harm"=>0.00016012965352274477,                                                                                
      "sexual"=>5.3972213208908215e-05,
      "sexual/minors"=>4.74193220725283e-05,
      "violence"=>0.06760840117931366,
      "violence/graphic"=>0.0018866335740312934},
    "flagged"=>true}]}
=> 
{"id"=>"modr-6bnr13tBOQd0EqfIWm6bKZX7t9Hy0",
 "model"=>"text-moderation-004",
 "results"=>
  [{"categories"=>
     {"hate"=>true,
      "hate/threatening"=>false,
      "self-harm"=>false,
      "sexual"=>false,
      "sexual/minors"=>false,
      "violence"=>false,
      "violence/graphic"=>false},
    "category_scores"=>
     {"hate"=>0.6628993153572083,
      "hate/threatening"=>0.0006026297342032194,
      "self-harm"=>0.00016012965352274477,
      "sexual"=>5.3972213208908215e-05,
      "sexual/minors"=>4.74193220725283e-05,
      "violence"=>0.06760840117931366,
      "violence/graphic"=>0.0018866335740312934},
    "flagged"=>true}]}