I had lots of fun playing with AI GPT-3 text generator for this episode. One thing I noticed about the AI is it appears to be trained on mainly US data. Whenever I asked it to introduce a person (with a name I made up) it always said they were from a state in the US, unless I specified which country they were from. When I specified another country the introduction was cliched - like surfing and sharks for Australia.
AI will be biased according to the data sets that it has been trained on by the humans that set it up. If you want to learn more about that Joy Buolamwini explains this beautifully in her TED talk here... https://www.ted.com/talks/joy_buolamwini_how_i_m_fighting_bias_in_algorithms?language=en
The youtube algorithm changed to be AI based about 5 years ago now - that is when all baking and cake content suddenly started declining in views and content-farm cooking videos took off. I wonder what bias exists in the youtube algorithm. My youtube partner manager of several years ago speculated that it favoured male voices because even the sections that used to be female dominated, like cooking and beauty, suddenly changed to have a surge of views to male presented channels. Only youtube itself has access to the data that would be needed to test that theory - it would be fascinating to look across the whole platform and see if that has any truth in it.
Sassy Mama
2022-04-27 12:03:08 +0000 UTC