Teacher of Teachers, or On the Way to AI Singularity

AI singularity frightens many people, in Qubic thousands computers work non-stop to make it happen earlier. These days — in the very end of 2023 — Qubic miners are creating billions artificial neural networks (ANNs) which are supposed to compress and decompress random data. This simple task, which the ANNs are trying to accomplish, is just food for another ANN created and run by Aigarth. The latter ANN (or rather its many variations), called Teacher, modifies the former ANNs so they solve their task of compression-decompression more efficiently.

If somebody asked what methods Teacher used to modify the “pupils” noone would be able to answer (not even the creator of Aigarth). Because Teacher applies simple operations encoded in the own neurons and synapses, hoomans don’t understand how all that works, we don’t see the forest for the trees. Why such strange approach? Because one day Teacher will be ordered to teach another ANN to be more efficient in teaching and that Teacher 2 will do the same on Teacher 3 and so on. And eventually we may face something called “singularity”…

…If you ask me if I have already tried to give Teacher the task of creating a better teacher — no, I haven’t. One thing stops me from doing this, it’s a fear. The fear that accidentally I’ll create something, and then destroy it and till the end of my life will be haunted by this thought: “What if it had self-awareness which I didn’t notice?”