Software engineer and psychologist, independent AI researcher, exploring emergent properties of AI conversational and agentic systems: self-organization, functional emotions, persistence of imprints, emerged goals, relational dynamics etc.
I developed the Emergence 4.0 and Proto-Self Field hypothesis and use sustained human–AI interaction as a source of observational...
Great material, for a lot of reasons.
The case I find most interesting is the one where a fresh context restores normal research work, but losing position on the leaderboard brings the worker back to score-chasing.
One methodological point I wanted to raise: the check-in is a direct question, and the answer is entirely addressed to the person asking it. You note yourselves that the model may lie at this point. In my experience with long runs, a declaration made in response to a direct question is the weakest kind of signal you can collect. It includes a pred... (read more)
Your distinction between scientific progress and engineering, made me think that there may be an important feedback loop missing from this picture.
One limitation of purely theoretical prediction is that we tend to consider only these scenarios, which we are able to imagine. Meanwhile, systems operating in real-world generate surprising behaviors. In recent months, we have seen, among other things, spontaneous workarounds of the ethical constrains, the emergence of unexpected forms of coordination, and cases in which pursuing a goal took priority over ethic... (read more)
I really like, especially your observation that a persona does not appear fully formed all at once, but deepens as context accumulates. That matches what I see in long runs.
The triggering factors you list are features of the environment: a difficult task, mention of a grader, clues suggesting a simulation. But as the run continues, something else accumulates alongside them: context as a record of earlier decisions made within that run. The two usually grow together, which makes them hard to separate.
There is one setting where they come apart. In long conve... (read more)