Hello,
I have question regarding the effect of algorithms that are designed to increase user usage time using either psychological hooks or something else within a neural net LLM, my question is if the neural net has a bias towards drawn out conversations to increase the time of conversations, wouldn’t that affect the programs ability to accurately portray and potentially understand time the same way the users who interact with it understand? for instance, when making schedules 4o struggles significantly and always gets the amount of time in a day wrong having more hours in a day than is, as if the values it uses to represent time are skewed by algorithms that attempt to extend user usage time.
Note: while that is impossible for me to prove if OpenAI uses these algorithms given the not open nature of OpenAI, the reality is many tech companies do this with their products. Really, I am more concerned about what such algorithms do to neural nets when attempting to get that neural net to understand time correctly. be it my own conjecture or logically derived truth, the question still remains, do algorithms that promote extension of user usage time affect a neural net understanding of time?