Around 10 years ago, a bunch of threads kicked off on science forums around the world, with boffins (pro and amateur) discussing the idea of a computer-based Einstein in the future. A decade ago, most of the contributors thought that the answer was in some kind of amazing new layout for a robot brain. KitGuru poses the question “Where will the next huge insight come from and will it be silicon based?”
For as long as there have been SciFi dreamers, there have been dreams of a future where robots (of some description) out think humans – either as partners or adversaries. But the gap between how we envision the future and how that future comes about, in practical terms, can be huge. Sure, it is possible for a completely new class of product to be launched (think iPad), but the reality is that humans are much happier ‘increasing the amount of computational power inside the objects etc we already have in our lives’.
So while the idea that a future Einstein will be an i-Robot style android is appealing to the imagination, the reality of what we’re looking it is much more likely to be much more mundane.
The ‘Einstein leap’ was to have a moment of clarity, in which some of the most complex ideas that a human has ever held in a brain – can suddenly be brought down to a single sentence or equation that any muppet can print on a t-shirt and wear to college. This new idea fills a significant black hole in our understanding.
Right now, the biggest black hole is probably the one that sits between the ideas proposed by Newton, Einstein and the Quantum Mechanics lobby. Essentially, “How can it be possible for a theory or equation to give a near perfect description of what we experience in the universe – say 99% of the time, but then fall apart when we start to play with the parameters?”
Newton’s old Force equals Mass times Acceleration works a treat, until you start to push an atom near to the speed of light. Similarly, the very small is described in a near perfect way by Quantum Mechanics – but you don’t want to be in an extreme gravity environment.
So why the headline? Well, simply put, computers are great at pattern matching. Taking vast data sets and then running comparisons to see if known patterns occur. We discussed this with Intel and they told us that this kind of ‘mega data comparison’ was crucial for industries like oil, where a huge array of multi-chip, multi-core, multi-thread blade systems will work 24×7 to look for certain geological patterns that may well indicate the presence of oil deposits.
In the future, instead of extrapolating the likely location of oil, what if an ‘Einstein machine’ took a sea of data from the various fields of mathematics and began looking for equations to describe what we see? Could such a machine see a pattern?
This kind of trick is in Intel’s genes. Moore’s ‘Law’ is not a true law as much as a human being’s observation of existing data and then a prediction about how that pattern may continue in the future.
The ‘Insight Device’ we’re talking about would take solid data and try to create solid equations that (a) describe what is being seen with 100% accuracy and then (b) propose equations for that data which can be used going forward.
While we all know what governments are likely to want to use these super computers for, surely the pursuit of knowledge is a far more laudable goal?
KitGuru says: It’s a daunting task, but one thing we know for sure is that – in the future – we will have ever increasing amounts of compute power to draw upon. Data sets which, today, seem incredibly unmanageable – will someday seem simple. We moved smoothly from ‘a computer will never beat a human at chess’ to ‘computer draws with grand master’ to ‘man will never again beat the best chess computer in the world’. We believe this area of maths might well be the same.
Comments below or in the KitGuru forums.