Distilling Free-Form Natural Laws from Experimental Data

For centuries, scientists have attempted to identify and document analytical laws that underlie physical phenomena in nature. Despite the prevalence of computing power, finding natural laws and their corresponding equations has resisted automation. A key challenge to finding analytic relationships automatically is defining algorithmically what makes a correlation in observed data important and insightful. We have developed a technique for extracting the laws of nature from experimental data by identifying invariant and conservation equations. We demonstrate this approach by automatically searching motion-tracking data captured from various physical systems, ranging from simple harmonic oscillators to chaotic double-pendula. Without any prior knowledge about physics, kinematics or geometry, the algorithm discovered Hamiltonians, Lagrangians, and other laws of geometric and momentum conservation. The discovery rate accelerated as laws found for simpler systems were used to bootstrap explanations for more complex systems, gradually uncovering the "alphabet" used to describe those systems.

For more information, visit http://ccsl.mae.cornell.edu

-

Category:

Science & Technology

Tags:

• CCSL
• Cornell University
• Automating
• Science
• AI
• Robot Scientist

[netsight]

It maybe possible to set it on a crusade to discover the Grand Unified Theory Though what it would discover maybe incredibly complex and would probably need algorithms to filter out readable data.

[stokepogue]

Can this algorithm teach someone to speak a little clearer?

[netsight]

This algorithm would aid in the development of natural language processing (NLP) machines too.

[Cubelarooso]

DO IT ON LOVE

[patsuloi]

Using angles to represent the local coordinates of the double pendulum implies specific geometric symmetry.

[wulf8121]

Damn! Wish i thought of it first!

[Zalo10]

DO IT ON THE STOCK MARKET.

[fahed79]

my head hurts.......

[TiagoTiagoT]

also it would probably be interesting to hear what the machine got to say about the distribution of prime numbers

and just for kicks, feed it with audio recordings of people talking in as many different languages as possible

oh, and give it as many genomes as you can get your hands on

[TiagoTiagoT]

among the things that i think would be interesting to feed this program are images and measurements of the Sun, Earth's seismological history (including volcanic activity and massive landslides, including submarine ones), stock market data, occurrence and distrubtion of significant political events across the world, and perhaps also the universe's background radiation (atmospheric climate would be too obvious of a choice)