This felt very shallow which may be because of the breadth of topics the authors tried to address. Data management has been building for years on the pillars of data quality, with a passion that is at times counterproductive (looking at you single ideal schema of truth) and I feel this failed to put enough emphasis on the gradual transition to new mechanics of trust as a counterweight to probabilistic answers. We are falling into the trap of imagining robotic brooms instead of vacuum cleaners. I don't see LLMs perfecting approaches with singular focus on precision but I do see them introducing new with focus on convenience (is that not what we are witnessing with evolution of search?)
replies(1):