Hans Moravec has suggested that the human brain has a processing capacity of 10 quadrillion instructions per second (10 billion MIPS). In comparison, it was announced today that the fastest supercomputer in the world, called Roadrunner and devised and built by engineers and scientists at I.B.M. and Los Alamos National Laboratory, is capable of handling 1.026 quadrillion calculations per second.
In 2007, it was announced the previous fastest supercomputer, IBM’s Blue Gene /l, had been upgraded to achieve 478 TFLOPS sustained and 596 TFLOPS peak. So in less that 12 months, we’ve doubled (hey, I should create a law around that prediction).
So, if Moore’s Law holds out:
2009 – 2 QIPS
2010 – 4 QIPS
2011 – 8 QIPS
2012 – 16 QIPS – which puts it 6 QIPS above the suggested ability of a human brain.
Are you ready for that? Do you think the human race is ready for that?
We have NO IDEA what the consequences of that are. On one hand, it could be nothing. On the other hand, what if sentience is nothing more than massive computation?
Either way, here we are, a mere 4 years before a machine is likely to be built which will have a bigger brain than a human and we aren’t even discussing what that means for the human race.
Well, that’s not exactly true – Tyler’s Singularity Institute are discussing it, but where is the debate in mainstream media, in the government, in polite society?
It reminds me of a chat I had with Australian SF author Damien Broderick over dinner about ten years ago. I asked him when he thought these subjects would be discussed by the general populace. He replied “when it’s way too late to do anything about it”.
Well considering that this new supercomputer “Roadrunner” is being put into use running simulations on ageing nuclear weapons to make sure they’ll still work … I think it doesn’t mean much for the human race. It pretty much just confirms that our number one priority is annihilating ourselves and the planet.
Assuming that some of these supercomputers survive being dragged off for classified military use then I think we’ll start to see some headway with collecting and processing quantities of data sufficient for simulating the future.
Once we start coming up with a language for describing the human mind and how people think then we can plug that into equations and accurately predict the decisions people will make. Apart from that there’ll be the natural stuff like accurate weather predictions months ahead, predicting natural disasters, ability to provide governments with far more accurate and useful economic data to inform decision-making, avoiding recessions and the such.
As far as a supercomputer with the processing power greater than a single human mind? I don’t know what that means … if we’re starting to go down the path of AI …
Nath, remember that most of the technological developments that have been made by the US in the last 60 years came from the industrial-military complex. That’s how the system works – they use taxpayers money to spend with private and public corporations to invent new technologies for the military, then the discoveries make their way into products which are then sold to the general public – at a massive profit. So the taxpayers fund the R&D and then have to pay the profit for the finished product. It’s almost criminal.
True true, so apply Moore’s Law … but for actual commercial application of the technology add 5 years? a2+5
Cam,
Do you ever wonder how come humans are the only species on earth who have evolved to this techinical stage? Our nearest relative, the chimp can only use primitive tools like sticks & stones and yet humans are supposed to be the newest species on earth. We haven’t been here all that long. Why hasn’t any other species on this planet developed as we have?
Not sure what it might mean for the human race so to speak. But I was thinking about your numbers in that you might have a computer that is capable of 16 QIPS but then you need some code monkey to write the lines of code to enable it to do that. Is that possible, I dunno to be honest. But I guess that assumes that it is calculating with one application at a time. The alternative is a few things: A computer capable of doing some impressive multi-tasking or just doing things like simulations for weather etc really quickly. In which case not that a big of an impact on the human race.
So my point is that unless we can write the billions of lines of code to execute what I think your suggesting, then it won’t be that amazing as far as the comparison to the human brain. Or… do we program them to program themselves… now we have a totally different scenario… Enter Terminator…
Gran Jan, yeah we have bigger brains. Well… some of us do.
Dave, you’re right that there will need do be some kind of software running on the machine. Perhaps we’ll ask the supercomputer to design it’s own code? Perhaps not when its at 16 QIPS, but what about 5 years later (2017) when its doing 1028 QIPS, ie is running ten times faster than a human brain?
The trick here is making code that learns. Our brains got us to where we are because the brain is incredibly freaking cool in that it can, on they fly, remap itself. You simply cannot remap a CPU on the fly. You can rewrite software on the fly, but its still not an issue of scale. The closest thing I think to AI these days are high level programming languages. You write a couple of lines in almost-english, the computer fills in the rest. Unfortunately, nothing particularly mind blowing has happened in this field in a good many years now, the current paradigms have been refined but theyre pretty close to as good as theyre going to get. I haven’t the faintest idea what the next evolution of it will be, and I suspect we wont see it until we are stumbling into the Quantum era.
I think you’re giving the brain way too much credit. There’s nothing mystical about it (at least that I’m aware of). I don’t think we have any idea what a supercomputer running 10 – 100 QIPS will be able to accomplish. Maybe you’re right and it will take a new kind of software, a new approach to programming altogether. That’s certainly been Jaron Lanier’s argument for a long time. Anyway, whether or not we end up with AI in the short term or not isn’t the point. The point is that by 2012, all things being equal, we’ll have built an electronic brain capable (at least in theory) of higher orders of pure computation than our own brains. In less than a century of computing, we’ll have built a machine capable of more computation than our own brains.
Now go watch Craig Ventner’s recent talk up on TED where he explains how his company is building new forms of life by cutting and pasting chromosomes.
We’re approaching the dawn of a new age.
Cam,
Didnt you notice in the graph I sent you that moores law doesnt hold true these days bacause we have reached the limits of the current technology. Its going to take a quantum leap in technology to something like nano-tubes (which by the way when injected into lab rats cause asbestosis like cancers) before we can move forward.
Besides this. I personally do not feel that floating point ops are enough. Its going to take more of an understanding of how the brain works BEFORE computers have a hope of thinking like we do.
when eventually they do manage it and we all become transhuman, what happens to art, music, the appreciation of a nice wine, enjoyment of where and who we are with. I think that these things will become less and less important antil they reach a point where they dont exist at all. At this point we bacome machines.
J
J
And yet the MIPS record for the supercomputers keeps doubling every year. And with PCs, we’re getting around the gate limits for chip design by building multiple cores, etc.
And who said computers need to think like we do? Perhaps they will think in a way unique to silicon brains.
What happens when we become transhuman (assuming that happen)? I don’t know mate. But I’m looking forward to finding out.
I think that processing capacity is only one small element in the human machine….
The stuff that really counts is the ‘irrational’ side of humans, and physical make up of humans, the ability for humans to change their physical make up through environmental exposure and practive… which counts.
But heck, what would i know, I’m just a marketer…
Steve.
Steve, do you think these things you mention are products of the human brain? Or something else?