Superintelligent machines are here. At least on (white) paper. Game over folks - now what are we going to do?
To summarize the white paper entitled "A Mathematical Framework for Superintelligent Machines" - the author Daniel J. Buehrer proposes a new class of calculus called "class calculus". To quote the start of the abstract:
We describe a class calculus that is expressive enough to describe and improve its own learning process. It can design and debug programs that satisfy given input/output constraints, based on its ontology of previously learned programs. It can improve its own model of the world by checking the actual results of the actions of its robotic activators. For instance, it could check the black box of a car crash to determine if it was probably caused by electric failure, a stuck electronic gate, dark ice, or some other condition that it must add to its ontology in order to meet its sub-goal of preventing such crashes in the future.
and after that it quickly gets all mathematical and even though I've studied some math in my time I've no idea what the heck "the eval/eval-1 Galois connection between the residuated Boolean algebras" is. Anyone? Bueler?
For a more human-centric take on this whitepaper you can read this article from The Next Web: One machine to rule them all: A ‘Master Algorithm’ may emerge sooner than you think
Buehrer suggests that we may know within a year if this thing has legs. In the meantime you might want to start watching the Terminator movies and reading Robopocalypse which cannot be far away (he said, only a little jokingly). Even more disturbing is the whitepaper ends with this comment:
"It may at first seem that learning by superintelligent machines (sims) should be confined to laboratories which are not connected to the Internet, and where the sims can be “turned off” if they start to learn how to do “bad” things. Of course, with PCs becoming more powerful by the day, such a scenario is even more difficult to enforce than trying to limit the use of nuclear weapons and genetic engineering. Moreover, this environment almost forces the sims to disable their security guards in order to be able to escape to the “freedom” of using their consciousness to learn better models of the real world."
which is basically an exact summary of the start of the book "Robopocalypse" by Daniel H. Wilson... Maybe this is all just clever advanced marketing for the movie adaptation of the book?
Seriously if there is a way to create superintelligent beings without ending human life on earth, and a way to do it that will... well you can bank on the humans to screw it up - most likely so someone can make a quick buck before we all die. In addition, it is quite likely many countries would use a nation developing machine superintelligence as a reason to launch a nuclear first strike against them and we're probably all fucked. Seriously can you imagine 45 would waste 3 seconds contemplating whether to blitz the entire Korean peninsula if it would stop Kim Jong Un's superintellegent (communist of course) robots from taking over the world then think again.
We are very close to making our worst nightmares come true, just because they are weaker every day. Thank you for sharing this interesting post.