Slashdot Mirror
Computer Prediction Used to Design Better Organic Semiconductors
An anonymous reader writes "Creating a flexible display requires finding an organic material that's both durable and capable of carrying an electric signal fast enough. To create such a material requires choosing the right compound and combining it with an organic base material. It's a hit and miss affair that can take years of synthesis to get right, but even then the final material may not be good enough. Stanford and Harvard researchers have come up with a much faster solution: use computer prediction to decide on the best compound before synthesizing begins. They also proved it works by developing a new organic semiconductor material 30x faster than the amorphous silicon used in LCDs."
I forgot that SAR means about twelve things, so for the context insensitive people, in this context it means Structure activity relationship
Jehovah be praised, Oracle was not selected
What's going on with Slashdot lately? There's a major advance in materials science research and so far a full 100% of the comments are either sarcasm about how stupid the research is or someone arguing that the solution is obvious.
I just... don't even know what to say anymore. Companies have been pouring money down the drain trying to discover these materials and these guys figured out how to do it more quickly, more effectively, and at a fraction of the cost. Obviously this is a problem that has been looked at by hundreds of engineers and scientists, and this is the first group to successfully apply this technique; and all we can muster is a collective circle jerk trying to sound smart.
Pathetic
Over time, there have been more and more ways of getting computers to our scientific work for us. There's been success with this sort of thing for a long time. For example, in math the mid 1990s the Robbins conjecture http://en.wikipedia.org/wiki/Robbins_algebra was proven using essentially an automated theorem proving system. Similarly, Simon Colton has done work with computer systems that can notice patterns and make novel mathematical definitions and conjectures without human input ahref=http://www.cs.uwaterloo.ca/journals/JIS/colton/joisol.htmlrel=url2html-7891http://www.cs.uwaterloo.ca/journals/JIS/colton/joisol.html>. There's been work in other fields as well such as in automated experiment systems in biochemistry http://www.aejournal.net/content/2/1/1.
One interesting thing about TFA is that it is suggesting that powerful enough systems might be able purely through simulation to predict what new compounds are worth investigating. In principle, this could lead eventually to self-improving systems where a system designs better and better algorithms and hardware for itself which it uses to then design even better ones and so on. This sort of Singularity scenario generally seems implausible to me but it is one of the more plausible Singularity scenarios and articles like this make me wonder if it should be taken seriously. Obviously, there are serious limits on that sort of thing, since one runs into the laws of physics and the laws of mathematics eventually. You can't improve algorithms beyond a certain point (you start running into P!=NP sort of issues). And you can't improve physical efficiency beyond the laws of physics. In essence, even if I can look through a search space a 1000 times as fast, that doesn't mean I will find a solution that is a 1000 times as good. But with access to both software and hardware improvements, substantial improvement may be possible.
Of course, at the current stage this is highly limited. The simulations in question in their current forms can at best target specific molecules that look promising. They are very far from actually predicting the exact behavior. Indeed, my old college roomate is a physicist who works on computational physics and related issues, and one thing he and a few others have been working on for a long time is trying to derive accurate behavior for water from quantum mechanical principles. This is still not doable beyond a rough approximation. The computations are simply too difficult. And as TFA notes, figuring out how to synthesize new compounds can still take years. So this is a really interesting result, but it isn't something that is likely to see immediate impact. Still very cool though.
Oh they're supposed to be Standfor and Hardvar? I've been saying them wrong all these years!
The enemies of Democracy are
though I do have to ask why it has taken so long.
Take a few years of organic chemistry courses, especially the labs, and you'll become astounded that they've gotten even this far.
Organic chemistry, compared to all other things considered "hard" science, is so difficult and the fundamentals so poorly understand that compared to physics or inorganic chemistry it might as well be mysticism. I struggled with it, trying to treat it like a science, for a whole year before my professor finally admitted that they don't know exactly how anything works; the core theories are good science, but have little more real-world proof than quantum physics. And succeeding at something novel in applied organic is far more art than science, despite the need for a post-doctoral scientific background.
They deserve serious credit for this kind of breakthrough, not questions about why it's "taken so long".