Many human minds can make insights and progress that a single human mind never could.

Unlike a dog, humans can make progress in a way which advances our whole base of knowledge forward.

From my perspective of thinking this through within a single organic human mind, it’s incredibly difficult to see any true limitation to what we could achieve by simply iterating on this process. Human minds are able to create abstract concepts and ideas, we can then use those ideas to make progress on newer, more complex ideas. What limitations would this kind of setup have? This question cuts to the heart of what separates us from dogs. How far do we think we could get with something like a trillion human-level intelligences-all networked together so they could quickly communicate and collaborate-running at a very high clock speed so that one minute of our time would be equivalent to a year of their time? When we imagine true superintelligence, we have to imagine that there is at least some task out there which could never be achieved by any arbitrarily high number of human minds given nearly infinite time.īefore we go too deep on hard superintelligence, let’s look at soft superintelligence. Is there any meaningful problem we could task this dog-brain network with, given unlimited time, where it would outperform even a single human mind?Īssuming the dog brains don’t spontaneously evolve human-like intelligence, no matter how many dog brains you set on a task and no matter how much time you gave them, they would not make progress on any task which required human-level intelligence. To crank up the example Vinge has given us, let’s not just imagine “a thousand years of doggy living,” instead, let’s imagine one billion simulated doggy minds running 1,000 times faster than an organic canine brain. When we talk about ASI (Artifical Superintelligence), many people fail to really define the term. Would a thousand years of doggy living add up to any human insight? Many speculations about superintelligence seem to be based on the weakly superhuman model. Imagine running a dog mind at very high speed. It's hard to say precisely what 'strong superhumanity' would be like, but the difference appears to be profound. 'Strong superhumanity' would be more than cranking up the clock speed on a human-equivalent mind. In Vernor Vinge’s 1993 essay on the technological singularity he gives the best (and simplest) distinction I’ve seen between the idea of a true superintelligence, and what I will refer from here on as soft superintelligence.

Diaspora egan