Very interesting interview. I learned something about the limitations of LLMs, and their development and applications. One constructive comment - I think Karin confuses “bias” in AI models with “unpleasant (and uncomfortable) results”. One needs to be honest about what the data shows, then be willing to address the underlying causes of those adverse outcomes rather than manipulating the data to distort the message. In the longer term, that is where “using AI for the good” will be most positively impactful.
Very interesting interview. I learned something about the limitations of LLMs, and their development and applications. One constructive comment - I think Karin confuses “bias” in AI models with “unpleasant (and uncomfortable) results”. One needs to be honest about what the data shows, then be willing to address the underlying causes of those adverse outcomes rather than manipulating the data to distort the message. In the longer term, that is where “using AI for the good” will be most positively impactful.
Thanks for your comment! It’s a very complex topic and it gets oversimplified for sure.