I think more likely we don't even estimate the right thing. We can't predict all of the effects of LLMs, or any technology that seems likely to get wide adoptio

I think more likely we don't even estimate the right thing. We can't predict all of the effects of LLMs, or any technology that seems likely to get wide adoption and use in many contexts. Look at smartphones. Who predicted a correlated rise in depression and anxiety among teen girls back in 2010 when the iPhone 4 (with the first front-facing camera) came out? Human history has many examples of technologies causing all kinds of unpredictable and unintended consequences, both good and bad. LLMs and the more advanced "AI" we can expect will almost certainly have the same problems and benefits. "AI" technology will get deployed by companies staking out new markets and niches, with little thought to consequences as they focus on market share and profits. Governments will find new ways to use and abuse LLMs and other machine learning tools (actually that already happened with facial recognition, to give one example). Debating the potential consequences may interest some people, but I would caution that, first, these "debates" won't affect corporate or government plans, and, second, we will almost certainly debate the wrong issues. History provides many examples, from the rise of agriculture leading to militarized nation-states and tooth decay to the Twitter "world town square" leading to political manipulation and providing an outlet for bigots. Pandora's Box already opened, and now any companies or governments that hesitate or show moral constraints will get left behind by those with fewer scruples. Look at what happened with "organic" food to see how companies react to even small threats to their market share.