> I think human intelligence hits an asymptote due to mortality. There's only so deep an individual can go, and a limit on how much you can replace depth with breadth.
I don't know exactly what you mean by "human intelligence", but in general we get deeper not by stretching vertically, but by moving vertically, if you follow the metaphor. Many software engineers today don't even have a passing familiarity with the deeper levels of the stack they're programming on, and that's all new stuff in the last ~50 years.
Very good point! To the extent that we can build on abstractions and black boxes we stand on the shoulders of giants. BTW, perhaps s/human intelligence/science & technology.
Does this really work about as well (e.g. leaky abstractions but better than nothing) in all disciplines? Learning math is a pretty much a 'climb the ladder' experience. You don't have to literally discover it all for yourself, but you do still have to learn it. I assume e.g. biology is similar; there are absolutely abstractions upon abstractions, but it takes a tremendous amount of work to level up to the point where you are making breakthroughs.
Certainly low hanging fruit abounds, but our individual capacity for learning seems fairly limited in the great scale of things, and has not advanced much in the past few centuries. As a species, we either gain an order-of-magnitude increase in our ability to learn/retain knowledge, or the computer does it all for us. Not sure how much I like the potential adverse side-effects of the 2nd case.
Evolution of science / technology is typically drawn on an exponential scale and we're staring up the side of a cliff ahead of us. There are a lot of marvels in the world we have built, but most of them to someone with 15 years of training in the relevant field are not black boxes or magic. I just imagine climbing much further up the cliff and you start to get to that point.
E.g. I was watching a video of how the Toyota Prius eCVT works last night. I know fuck-all about CVTs, but I have a passing familiarity with planetary gears, and could follow the 15 minute explanation / walkthrough and learned enough to distinguish eCVT from magic. In 100-500 years I would be surprised if the underlying technology of our personal transportation devices would be remotely so 'accessible'.
Isn't it only a matter of time before computers are designing and replicating devices that are sufficiently beyond our capability to understand how they even operate? Or synthetic compounds created by computer algorithms that cure disease but we have literally no idea the mechanism of action? Maybe in a very few problem domains today this is already the case.
Great thoughts (in this and your other comments here). I have to think that for better or worse, even if human lifespans are drastically increased, it is inevitable that we will reasonably quickly reach the point where computers essentially know more than we do. 500 years might give us time to climb a much higher ladder of knowledge, but computers don't have to climb it at all; they will all have access to all the knowledge, effectively immediately. Also, a computer will have access to extreme depth and breadth of knowledge, which would be impossible for a human, and which could lead to all kinds of cross-discipline breakthroughs.
Of course there's also the middle ground option, where the clear distinction between human and computer becomes blurred, and eventually meaningless.
On the other hand, there are physical limits to computing power. Computers will always be able to solve simple problems faster, but I suppose it isn't a given that they will be able to solve complicated problems better, than humans. (This other article on the homepage got me thinking about this more: http://www.scottaaronson.com/writings/bignumbers.html)
I often think about how humans manage to build upon previous knowledge, and how much more difficult it gets to add another abstract layer, but the point* you just made about computers designing increasingly complex paradigms really got me thinking/scared!
*
> Isn't it only a matter of time before computers are designing and replicating devices that are sufficiently beyond our capability to understand how they even operate? Or synthetic compounds created by computer algorithms that cure disease but we have literally no idea the mechanism of action? Maybe in a very few problem domains today this is already the case.
I don't know exactly what you mean by "human intelligence", but in general we get deeper not by stretching vertically, but by moving vertically, if you follow the metaphor. Many software engineers today don't even have a passing familiarity with the deeper levels of the stack they're programming on, and that's all new stuff in the last ~50 years.