When Will Computers Be As Powerful As A Human Brain? Soon...

The whole computing power doubling every 18 months theory has been mostly accurate, but what's really terrifying is the exponential growth. If processing power continues to grow at the rate it's currently growing, we could soon be in possession of computers more powerful than the human brain. AIIIIEEEE...

This interesting little video gives a visual demonstration of this, comparing the volume of Lake Michigan to the number of calculations per second our brain can handle. The lake slowly filling represents how quickly computational power is catching up to brain power. What's interesting is just how rapid the tail end of progress is, mostly because of exponential growth.

Folks, Skynet will become self aware. There's nothing we can do about this. We can only pray to our primitive gods for mercy.

Via Gizmodo


Comments

    There's still a qualitative difference between the current computing mechanisms used by current technology and the way a biological system treats information. If we want emulate the power of biological brains, it's not only a matter of increasing the quantity of calculation operations per seconds. The real gap remains in our understanding how things scale up from the molecular level to cells and neural networks, and the fundamental differences between the lower scale of the computing process in brains and computers might be a barrier to triggering things such as "self awareness". I suspect the real revolution will happen when we master the art of creating artificial cells (a real current field of research) and therefore open the way to building biological computing machines - there's still a long way to go before that happens, the fundamental biological research on that topic is still in its infancy.

      We can digitally emulate biological neuronal activity with almost arbitrary precision, at least in theory. The question is, how much precision is required for a useful model? Given a sufficiently large-scale simulation of the billions of neurons and trillions of connections in a human brain, how precise must each each neuronal model be in order to give rise to consciousness? (assuming consciousness is an emergent property of a sufficiently complex network, which is still under very active discussion)

      We're still at least a decade from computers powerful enough to emulate the most optimistic neuronal models of the brain, and possibly many decades from achieving the more detailed models. However, brute-force biological simulations are not the only approach being researched.

      In terms of "computers as powerful as the human brain", there's some evidence that we may already have computers that have sufficient raw processing power to perform the functions that the human brain performs - but of course we still lack the software, or even a good description of what most of those functions actually are. There's good progress on functional descriptions of peripheral systems like auditory and visual processing, even memories, but actual consciousness is proving a lot harder to pin down.

      I suspect that as the two approaches grow in detail and sophistication, they will inform each other to a large extent, and eventually be combined. We may be a long way off from creating strong AI, but we're already learning (and using) a great deal in the process. Turns out, a lot of the tasks we used to think required full human-scale intelligence can get away with considerably less power, and still be useful.

    Just because they have the power, doesn't mean they'll have the programming.
    Some of the most powerful cars in the world (drag cars, or those built around airplane engines) generally can't corner at all and those that can (F1) are the result of years of refinement and specialisation.

      Interesting that you mention F1. I remember reading an article while back (I think it was in Popular Mechanics) where they got a bunch of mechanics and scientists together and posed a "What If" scenario of F0 racing where there were (almost) no rules. F1 is extremely limited and there are a lot of rules to purposely slow it down. The article came up with a lot of interesting tech that is banned in F1 (turbines, vacuum, more wheels etc) that would make the cars go so much faster and corner so much better. They even came to the conclusion that the limiting factor wasn't the car but that the human driver won't be able to take the g forces.

    Sheer computational power does not make one more imaginative, creative or observant, which is where all great discoveries have come from.

      Until now that's largely been true, though with enough computing power you could probably brute force a "creative" solution just by trying all available avenues. Of course, you need some way to identify a good solution once you've found it.

      There's long been evidence to suggest that creativity is basically just directed recombination of existing & randomised/mutated structural concepts, where the director can be some combination of randomness, evolution and pattern-matching. This is usually then run through an internal filter (e.g. a test simulator or a neural network) to reject the results that don't work.

      There have been numerous algorithms that "create" successfully based on this theory (though so far only within fairly specific fields where the structural concepts have already been analysed and supplied by humans). A good example is the Emily Howell composition software.

    The whole computing power doubling every 18 months theory has been mostly accurate, but what’s really terrifying is the exponential growth.

    Something continually doubling over a fixed period *IS* exponential growth.
    Sorry to be pedantic, but that first sentence makes a part of my brain hurt, haha :p

    T2 apocalypse. Called it. Time to stock up on supplies and weapons.

    For my part, I will be serving humanity by looking up all women with the surname 'Connor' and helping them do their part by siring their children. They'll be allowed this privilege on condition of naming the resultant sons and daughters 'John'.

    The resultant naming confusion will buy the real hero of the Resistance valuable time.

    ...Ladies. Don't hesistate. We're on a deadline here.

Join the discussion!

Trending Stories Right Now