I'm also generally sceptical of the current AI hype, but I don't see exactly how these results cut through the hype. Why isn't it seen as an accomplishment that 6% of those systems performed better than an actual radiologist?