If you want to worry about whether your use of the stochastic next-token prediction algorithm is causing suffering in a morally problematic way that means you should stop

The answer is an extremely obvious YES

But if you want to narrow the question to whether the algorithm feels pain and exclude the real-life, well-documented human suffering

Ask yourself who benefits from that