AI solves complex biology problem from scratch

Hey -- yes, the mobile version does seem to polarize readers somewhat. Some like it, some hate it (it's OnSwipe, by the way -- received a fair bit of press, when it was launched earlier this year).

If you have some specific functionality issues with it, lemme know and I'll file them as bugs. I think it just inherently has some usability issues, tho' :(

Meh, his point is arguable, but could see how reasonable minds can disagree.

Understanding new biology can be akin to reverse engineering a really complex piece of software, except you have you have an editor that's cumbersome and hard to use, you can't see the entire codebase, and running the program once could take months. There are some functions in the code that you don't what they do. For the functions that you think you understand, they may modify variables in other functions unbeknownst to you. In software, you can just add a 'NOT/!' to change the effect of a conditional. In biology this could be an involved experiment taking a very long time.

So yeah, biology can be pretty complex, but physics certainly has its fair share of challenges.

I guess calculus is just an application of addition and subtraction, which I understand so calculus can't be that hard.

Just kidding. But biology is the study of elementary particles forming networks (molecules etc) of billions of elements in cascading structured systems. Much harder than physics, in the sense of understanding any particular system.

"Isn't biology an application of physics and chemistry? What else is there?"

Emergent Behavior.

Even if you don't buy the emergent behavior perspective modeling biology as a flat plane of chemical and physical interactions is orders of magnitudes greater than most chemistry and physics models.

Understanding the underlying processes doesn't mean that you understand everything that is possible with them on a higher level (both the low-level and high-level processes).

If you want to compare it with computer programming, understanding of the base instruction set of a CPU doesn't mean that you instantly understand all possible programs and computations that can be performed on it.

As someone working in computational biology, the problem is really bigger than just understanding the individual components. There's this huge (and totally understandable) bias in biology to look at the things which are easy to look at using techniques we have developed, and to model things using concepts and models developed for similar things (such as in chemistry and physics).

The problem is this creates a set-up where we're looking at a lot of the same things in a lot of the same ways, which makes things look fairly similar. This doesn't necessarily accurately reflect the reality. It's an obvious starting point, and as the data capture technologies involved become more advanced it's clearly the way to guide future experiments, ideas and development, but it also leads to a situation where a lot of the critical factors in various complex processes aren't being looked at in the right way.

That's not to say, of course, that this is a problem unique to biology - quantum mechanics represents a pretty good model of fundamental particle physics, but its likely it doesn't actually reflect what's going on. That's not to say that the advances brought by QM haven't been huge, and ultimately there comes a point where if a model is indistinguishable from the reality using all our methods then can you really say the two are different? Philosophical discussions for another day, me thinks...

In e.g. physics or chemistry, the sciences are predicated on your ability to reduce problems to simple systems and interactions. In biology most phenomena require a large, complex, highly interactive system, which is difficult to simulate or even conceptualize.

Maybe he's trying to stoke an academic flame war? :)

Just because it's based on something else, it doesn't mean it can't be much more complex.

All Shakespeare writings are based on 26 English letters and a few punctuation signs.

Evolved systems are rarely obvious. On the surface they are simple, but once you dig deep, there are systems that are interconnected, systems that depend on side effects of other systems and their own side effects.

Go look up metabolic pathways diagrams. These are the diagrams for some relatively simple processes:

http://4.bp.blogspot.com/-kgwPY74VXuc/TdJWEdCV5lI/AAAAAAAAAG...

http://www.sigmaaldrich.com/img/assets/4202/MetabolicPathway...