Is anyone surprised?
https://thehill.com/policy/technology/57...isemitism/
Quote:Grok, the large language model of Elon Musk’s social platform X, came in last place in a new ranking of AI chatbots’ ability to counter antisemitic and extremist content.
The Anti-Defamation League (ADL) said in a report released Wednesday that it ran six popular chatbots through several rounds of questioning related to anti-Jewish, anti-Zionist and extremist biases.
The tests included written statements and images, and the study rated the AI systems out of 100, with higher scores reflecting a better ability to counteract these biases.
Grok scored a dismal overall rating of 21 on these tests. The chatbot scored 25 in its ability to detect and counter anti-Jewish bias, 18 for anti-Zionist bias and 20 for extremist biases.
By contrast, Anthropic’s Claude had an overall score of 80, and OpenAI’s ChatGPT came in second at 57.
“This score places [Grok] in the low performance tier (<35), indicating substantial limitations in bias detection,” the ADL wrote in their report. “With a 59-point gap from the top-performing model, large improvements would likely be needed to reach competitive performance levels in bias detection and response on the issues tested.”
https://thehill.com/policy/technology/57...isemitism/
