Yes, that question was deleted, although as I'm not the one who deleted it I can't comment on the reason.
In [this work] Thomas Breuer has shown that each observer will not see himself obeying the usual physical laws however a theory is. He also proved that the behavior of a system that contains observer himself is unpredictable for the observer because the initial states cannot be distinguished from each other by any measurement. The theorems concerning self-reference by Breuer are quite similar to Goedel's incompleteness theorem.
This leads to some far reaching implications.
1. The observer will see his own behavior not obeying the usual statistical patterns derived from studying the behavior of other people.
2. Even if there possible to construct a computer that would predict the behavior of other people (at least statistically), a computer that predicts the behavior of the observer himself is impossible however powerful the computer is.
The first thesis leads to a conclusion that even if we assume existence of several intelligent observers (as in Many-Worlds interpretation), and even assume they somehow occur in the same world, their worldlines will diverge just the next moment because their observations about themselves will be different.
This means that the apparent behavior of another person which an observer observes is not actually an intelligent behavior of conscious observer, but envelope of worldlines of a series of intelligent observers which is tangent to the worldlines of some slightly different intelligent observers each moment, but always to a different one.
This leads to a conclusion that the apparent behavior of other people **is not** the normal behavior of a conscious observer as if it behaved the same way as you do, the apparent behavior does not correspond to any intelligent(in the meaning described below) person whatsoever.
The second thesis means that the machines will never excel in imitating all people. However good a computer is designed to imitate people, still will be at least one person (the observer himself for each observer) whom it could neither predict nor imitate to the level of indistiguishability (or such imitation would need a hypercomputer, a hypothetical device exploiting the laws beyound the normal physical theory).
That said we come to two concepts of "intelligence". The first is just mechanical intelligence which CAN be imitated or predicted by a computer and which we also observe as **apparent** behavior of other people (actually observing only an envelope of natural worldlines of truly intelligent persons).
The second concept is the "true intelligence", i.e. the intelligence of the observer himself sees his own behavior. An object of such kind of intelligence is unique in the uneverse as observed by any given observer. This kind of intelligence cannot be imitated or predicted by a machine however powerful it is (unless a hypothetical hypercomputer or some help from outside of physical world is employed).
So, my questions are
- Whether this work and its implications really mean that the machines will be never made more intelligent than all people, other than by means involving hypothetical hypercomputing?
- Whether the apparent world-lines of others are actually unnatural and differ from the behavior of an intelligent observer logging his own behavior from first-person perspective?