In the opening paragraph of Die Frage nach der Technik, Martin Heidegger describes science and technology in the following sequel "Questioning builds a way. [ ] The way is one of thinking. All ways of thinking, more or less perceptibly, lead through language in a manner that is extraordinary." For a long time I have liked this description as particularly fitting for the mathematical and physical sciences (and I did not read the rest of the book because I found it incomprehensible). These three sentences were quite clear - questioning builds a way towards understanding and understanding, miraculous and extraordinary, leads to discovery and innovation. This scheme lays foundation for the modern scientific method where the branches of the Tree of Knowledge represent the ways of a conscious mind.
So is there another way to obtain knowledge? one that is not preceded by understanding but where the existing knowledge is leveraged to move between points where a human mind cannot follow? Perhaps, if machines are involved.
One possibility is offered by the Data Science and it is fairly simple. We can gather colossal amounts of data and extract knowledge from them. Neither task can be done without sophisticated tools. Consider Large Synoptic Survey Telescope - it is a marvel of modern astronomy that (when operational) will collect about 30 terabytes of data per night. This data will contain information about planets, supernovas, dark energy, and much more, and this is all possible because the Universe observed at this scale appears to be quite simple.The fact that we can simultaneously track a relatively small number of variables on a huge number of objects through frequent snapshots is all it takes to make discoveries that thirty years ago were not thought to be possible. Another such tool is the Kepler planet finder telescope. With the marginal cost of finding new planets well below 1M a piece, commercial applications become possible…
When the setting is more complicated simple statistical tools tend to fail. Consider the Human Genome Project that was supposed to provide critical advances in medicine. Unlike in astronomy, the number or variables is very large and the data set is not big enough. In the end, fundamental advances in statistical reasoning are necessary.
Computational Science is another alternative, and the premise is simple - use the existing knowledge to compute and model the nature. In many cases it is the only way to gain some understanding of how the nature works. However, the methodological underpinnings of this process are inherently shaky, and when the stakes are high the results are questioned. Whether modeling of the climate or designing nuclear weapons the conclusions from computational models are difficult to verify and open to criticism. No wonder, even when only four arithmetical operations are concerned, the computer implementation may be deeply flawed as this incident shows.
For most of the other things, computational methods are very successful. Doron Zeilberger wrote:
Don't Ask: What Can The Computer do for ME?, But Rather: What CAN I do for the COMPUTER? This is a Faustian bargain but hopefully it will take some time before we get there. Nevermind, consider Inverse Symbolic Calculator and play with it a little. As this article illustrates the computer sometimes knows more math than you do. Other examples abound - consider Eureqa or software developed by Ayasdi. Eureqa will "discover" Newton's Second Law of Motion based on the results of a countertop experiment, and Ayasdi's software is likely to surprise you as well. Machine learning algorithms such as automatic translation between languages, Google's Page rank's calculations are all well known commercial applications where a great deal of computational power goes to produce results that cannot be in any sense verified by a paper and pencil calculation.
So in the end do not think that the battle is over and all goes to Computational Data Science and Engineering, Large Hadron Collider and other projects that need electricity but no coffee. Enjoy this and cheer up for human brain that operates on 1 watt of green energy.
No comments:
Post a Comment