Tag: Training data
Tag: Training data
-
Once upon a time there was transparency
In the age of neural networks, the transparency and explainability of AI models have become problematic issues. Transparency refers to understanding the model's processes and logics. Transparency includes having access to the training data, knowledge of the architecture, and information about testing and monitoring. Explainability is about the ability to determine semantically what caused the…
-
Does artificial intelligence present an opportunity or further obstacle to overcoming gender stereotypes?
Artificial Intelligence (AI) could both overcome and perpetuate gender stereotypes. Studies show that many AI systems are inherently influenced by gender and racial biases, reflecting the training data. Negative experiences on social media such as cyberstalking and revenge porn highlight malicious use of these technologies. The presence of virtual assistants with female voices and names…