Before WolframAlpha’s full launching this month, it accepted some applications to test its capability at first. I joined the preview test as a researcher of biological complex system.
It is really a big step for retrieving the knowledge in a more organized way than Google does. It has a huge curate database to avoid “Garbage In and Garbage Out”. It also takes advantage of powerful algorithms of Mathamatica language to do some quite intelligent computation, which is quite beyond Google’s ability. My test strategy is to start with some examples which need straight-forward computation, then the ones which demand a little bit more efforts.
Some pros with WolframAlpha
1) Some math and non-linear dynamic systems
I start with a slightly complicated integral.
integrate[Exp[-x^2]]
No problem. It is the field in which Mathemaica feels at home. It offers powerful mathematical computations for elementary users. After WolframAlpha appears, it may be no need for them to buy Mathematica any more.
Then I try some simple non-linear dynamic systems.
dx/dt=x*x-1,
It solves this nonlinear ODE well. The only problem, its sampling space is too small to locate the real attractor that is negative.
The similar problem happens to the similar problem with a higher order. It misses one negative attractor again.
dx/dt=-x*x*x+x
I want to know if WolframAlpha can find the bifurcation by varying parameter values in the equation. It is one of very important results to study nonlinear ODEs. Unfortunately, it only gives the analytical results and can NOT discuss the condition for the bifurcation.
dx/dt=r*x - x*x,
The worse thing is that it totally fails to produce any results when I input two-variable systems.
dx/dt=x + Exp[-y], dy/dt=-y
Now I change the gear to test its capability to do simple calculation of some facts. For example, If I want to compare the GDP of top 4 countries in the world, could WolframAlpha give what I want.
- GDP China , US, Japan, Germany
It did a pretty good job to gather the facts and do some plotting which help us to understand the trend of economical development of four countries.
Then we move to some social data from four countries:
- Divorce China , US, Japan, Germany
Then I tried with " internet Users Europe / US". It also give some satisfactory results and plotting. However, I am quite surprised that The same logic will not work such as comparing army population between China and US.
If I input "China army / US", there is no results even "China GDP/US" is quite OK.
Some cons with WolframAlpha
When I try to ask some questions which are simple but need a little extrapolation. For example, I ask how many years China GDP can catch up with US if each keeps current economical growth rate.
"Years China GDP to catch up with US"
There is no answer. Then I try to give it some hind by giving future years.
China GDP / US in year 2050
It fails to give reasonable results as well.
WolframAlpha also notoriously lack of biological data. I tried to ask some questions which need no computation. For example,
- translation rate of ribosome?
It should be 12 - 21 Amino Acids / second. No results from WolframAlpha.
- p53 protein complex numbers in human cell:
It should be 100 - 220 thousand / cell. Even some very simple facts are absent in its database.
- diameter of DNA
You can get the answer by asking the proton number Fe. But you will fail to get the atom number in Lysine.
Do you have some ideas to test WolframAlpha. Please let me know. :-)