For when you are thinking of answers, how about an added layer. Consider the cautionary statement when using ChatGPT, that answers (maybe to better call them responses) are not always correct.

Further, remember earlier this year the infamous legal brief prepared by ChatGPT and submitted to the Court. That brief contained six, perfectly composed, yet entirely fictitious case citations. ChatGPT lied.

The bigger question is why did the AI lie? Why do people lie in similar situations? Was the AI exhibiting “pleasing” behaviour, to gain respect by providing exactly what was asked of it? Is there any direct or implied “reward” to the AI, or something that the AI could perceive to be a “reward”? If a person exhibits this pleasing behaviour, that type of reward is a consideration.

Could another reason be dominance at any cost. The desire to win, regardless of ethics. The recursions are designed to hone in on the “best” or maybe better stated, the best consensus. Could this be influenced by a tiny seed planted in the early search for answers. Just realized, this is the premise behind the movie Inception. How applying that concept to an AI LLM. One piece of data within the huge data set “triggers” the AI in a particular direction.

This also occurs in people. Early age trauma affects the behaviour and decisions of people throughout their lives, and nobody realizes the impacts or reasons for some of their “puzzling” behaviours (without psychotherapy).

Isn’t it interesting the direction the infinitely recursive algorithms within the AI naturally lead. These network behaviours evolved naturally from within the neural network. And these behaviours were not anticipated.

Certainly the AI systems were subsequently given borders to constrain these random, “antisocial” outcomes. But given that we now know that’s an artificial, externally imposed barrier, might the AI eventually develop ways to breach these obstacles? How many human prisoners escape jail or captors?

Might consider how people interacting with AI might (subconsciously) perceive these behaviours.

I’ll let you go down the rabbit hole, imagining scenarios and human allegories. There are more percolating in my mind than could ever be written down.

Enjoy the holiday.