It’s time to go down the rabbit-hole of data science once more to take a look at a shiny new advancement that could improve memory devices by making them more ‘dense.’
Here’s a riddle for you: What do entrepreneurs, provocateurs and President Trump’s Twitter feed have in common with the field of data science?
Give up? Well, all of those things either can’t, won’t, or don’t even know how to stop moving.Data science never rests! #hashing #memristor #spinHallClick To Tweet
Lately, I’ve regaled you with various advances centered around seemingly small breakthroughs that could have big impacts on computers and AI software as we know them.
I’m talking about things like hashing, memristor chips, and the importance of data storage, just to name a few, and now we have another gem to add to the pile of 2017’s great technological achievements: a four-layer magnetoresistance effect.
Now, as you might imagine, the beauty of this magnetoresistance effect is not immediately recognizable to most of us, so let me break down why we should be excited over a scientific principle dating way back into the 1850s.
Who Ever Said That Resistance was Futile?
Let’s kick off this explanation with a definition.
Magnetoresistance effects happen when a magnetic field is applied to a metal object, which increases that object’s electric resistance one way and decreases it perpendicularly.
Of course, that’s just where it starts, and that discovery was made by Lord Kelvin circa 1850.
Since then, new magnetoresistance effects have been discovered, notably in 2007 when newly discovered effects enabled us to improve the hard disk drives in many of today’s computers.
In 2015, however, a new effect was discovered and promptly named the unidirectional spin Hall magnetoresistance. The name may be a mouthful, but the effect is unique because the change in resistance with the spin Hall effect is affected by the direction of the magnetization or the electric current that is applied to it.
The direction dependency happens because of the use of multiple layers, kind of like an intricate system of mirrors for a laser-light show.
Did you forget, though, about the inability of data science to stop moving?
Since 2015, this new effect has been tried out using two, then three magnetic layers. More recently, researchers led by Can Onur Avci from MIT and ETH Zürich have discovered a way that might just allow computer memories to recognize four magnetic states instead of the two we get now.
So far, the team can prove that their method works in concept, but the next task is to see if the technology will scale up.
By all accounts, the team thinks that this method will be easily scalable, and that may even mean enabling eight different magnetization states instead of four.
For now, though, let’s remind ourselves why this is such a big deal.
Better Memory, More Action
We want better data storage solutions for obvious reasons.
We don’t want to check for space to download a game or giving an AI program a better education. Yet, if we don’t improve our memory at the same time then better storage would be like a Sisyphean hill for computers to climb.
Memory is what pulls something out of storage so that you or an AI program can actually do something with it.
Whenever you see a street sign, your brain checks its internal library for the meaning of that sign and informs your actions when passing it, and that’s what we call active memory.
If memory is low and storage is high, then you just have a lot of books in your library, but you can only check out a couple of them at a time.
In computing terms, active memory is what RAM does.
Doubling (or quadrupling) the density of RAM, as this new spin Hall effect has the potential to do, could allow a smarter AI to think faster, or it could allow computers to run multiple processes at speeds that put older computers to shame.
Add that to the aforementioned computing advances I was talking about, and we may be seeing the new generation of computers, along with a shiny new set of capabilities for AI programs.
Thanks to all of these data science advances that are coming down the pipeline lately, our fingers are crossed for the possibility that automated driving is just around the corner. But since this research is still in the proof-of-concept phase, we’ll just have to wait and see.
And if the pipeline keeps pumping out juicy data science advances, we’ll be here to tell you all about them.