This seems almost too good to be true.
ScienceDaily (Apr. 3, 2009) — If Isaac Newton had access to a supercomputer, he'd have had it watch apples fall – and let it figure out the physical matters. But the computer would have needed to run an algorithm, just developed by Cornell researchers, which can derive natural laws from observed data.
 |
| The Design Matrix: A Consilience of Clues by Mike Gene |
 |
| Your Inner Fish: A Journey into the 3.5-Billion-Year History of the Human Body |
 |
| Catalyzing Inquiry at the Interface of Computing and Biology |
 |
| System Modeling in Cellular Biology: From Concepts to Nuts and Bolts |
 |
| The Plausibility of Life By Marc W. Kirschner and John C. Gerhart |
 |
| Agents Under Fire by Angus Menuge |
 |
| Life's Solution by Simon Conway Morris |
 |
| Information Theory, Evolution and the Origin of Life by Hubert P. Yockey |
 |
| The Fifth Miracle by Paul Davies |
 |
| Nature, Design, and Science by Del Ratzsch |
 |
| Origination of Organismal Form by Muller & Newman |
 |
| Biased Embryos and Evolution by Wallace Arthur |
 |
| Rare Earth by Peter Ward and Donald Brownlee |
 |
| The Privileged Planet by Guillermo Gonzalez and Jay Richards |
 |
| The Way of the Cell by Franklin Harold |
 |
| The Volitional Brain by Benjamin Libet |
 |
| Evolution in Four Dimensions by Eva Jablonka & Marion Lamb |
 |
| The Evolution-Creation Struggle by Michael Ruse |
April 4th, 2009 at 8:13 pm
Hi Bradford,
You wrote…
I suspect it is true, but not as "good" as it sounds.
Just feed the computer some results from Quantum Mechanical experiments and see what happens.
We wouldn't even have to get that fancy. A simple three body problem would make it go "tilt".
Comment by Thought Provoker — April 4, 2009 @ 8:13 pm
April 4th, 2009 at 8:15 pm
10 PRINT We need more funding.
20 GOTO 10
Hey look, I made scores of scientists, bureaucrats and lobbyists redundant.
Comment by nullasalus — April 4, 2009 @ 8:15 pm
April 4th, 2009 at 11:11 pm
Well, this computer does not "derive natural laws from observed data," which would be too vague a task for a dumb computer. Rather, it automates a search for conserved quantities (invariants). That's useful when one analyzes the dynamics of a new physical system.
Still, there is a long, winding from identifying conserved quantities (assuming they exist) to building a new physical theory.
Comment by olegt — April 4, 2009 @ 11:11 pm
April 5th, 2009 at 5:05 am
Sounds like data mining to me. Used a lot in (e.g.) bioinformatics to discover patterns ("invariants" if you like) in gene expression profiles.
Comment by Raevmo — April 5, 2009 @ 5:05 am
April 5th, 2009 at 4:44 pm
I favor the "intelligence is algorithmic" viewpoint. If this viewpoint is correct than a computer AI of sufficient complexity should be able to do anything a human mind can. Much like a human brain you'd have to start with a useful set of toolkits built-in and the ability to detect invariants sure seems like a useful tool. But other than that this tool seems to have nothing to do with AI or with "understanding" the data. So I'd say this is impressive but in no way revolutionary.
Comment by Todd Berkebile — April 5, 2009 @ 4:44 pm
April 5th, 2009 at 5:01 pm
Two things:
1)
"The researchers point out that the computer evolves these laws without any prior knowledge of physics, kinematics or geometry."
"Their process begins by taking the derivatives of every variable observed with respect to every other"
These two statements are contradictory. To say that this could be used by Isaac Newton is silly – it was Isaac Newton who created derivatives – that is precisely what he discovered. Similarly, for physics that, in the future, might require new math, this will be a useless tool.
2)
TP pointed out – the really interesting problems simply involve too much data to be arrived at in this way, even if they are achievable using calculus.
Comment by johnnyb — April 5, 2009 @ 5:01 pm
April 5th, 2009 at 5:02 pm
Hi Todd,
Now you have gone and done it.
You have started another round of the Great Zombie Debate!
I will spare everyone the volumes of ink spilt on the subject by not including the multiple links I was tempted to include and only provide one.
In an attempt to stay topical, it is my position that Penrose's argument against "Strong AI" procludes the "intelligence is algorithmic" presumption.
I maintain that the only way AI can have consciousness is if it includes Quantum Mechanics (which is not algorithmic).
Comment by Thought Provoker — April 5, 2009 @ 5:02 pm
April 5th, 2009 at 6:10 pm
I suspect TP is right about this.
Comment by Bradford — April 5, 2009 @ 6:10 pm
April 6th, 2009 at 1:13 am
Sounds like a modified genetic algorithm. A neural network applied to the problem would also be interesting.
Off topic, but for anybody who wants to watch a recent Evolutionism versus creation science debate, you can see it at Evolutionism versus Creation Science debate.
Comment by William Wallace — April 6, 2009 @ 1:13 am