Slashdot Mirror
Scientists Invent Scientist
An anonymous reader writes "From the Boston Globe: 'Researchers said yesterday that they have created the world's first robotic scientist, a system that can form theories, devise experiments, and then carry out the experiments almost entirely without human help.' Now, if it could file patents and lawsuits, it would be ready to enter today's world of technology."
It's true that we have gone from doubling our knowledge of the world in three years to just eighteen months. NASA has data that is deteriating before it can be analized, so I think the following conderns are unfounded:
Some scientists questioned whether the system, dubbed the "Robot Scientist" by its creators, deserved the title of scientist. For human scientists, some of the most interesting discoveries happen when researchers notice something they weren't looking for and suddenly change course...
I think there is plenty of accumulated data that just needs basic analysis.
It's really interesting to think about this system and IBM's new Webcrawler in terms of AI though, and what we might accomplish in the next ten years.
Can I bum a sig?
...and the robotic scientist creates a better robotic scientist and so on and so forth...
This has "Escher drawing" written all over it.
Just as every college student has suspected at one time or another -- a machine could be doing their homework for them, and they could be doing something interesting instead.
-----------------------
You are what you think.
People in AI have worked on automated scientific discovery for decades, and some of their systems have also had robotic components. This seems like a tweak and a good sales job, not a breakthrough.
The robot scientist will have its work checked over by real people anyway, prior to any publication. It is likely to ignore any interesting, but narrowly irrelevant data and so could miss important discoveries. It can only work in fields where any underlying biological phenomenon is simple eg biochemical metabolic pathways. Many experiments contradict each other, where the underlying biology is extremely complex, with a host of competing factors and extremely sensitive to slight changes in experimental reagents. I'm a scientist, and I'm not too worried. Modern maths uses number crunchers too, like with the 3 colour map problem, but the proofs are always checked over. I guess the difference is these maths problems would take so long in human hours as the dissuade anyone from starting. This isn't the case in most of biology so I reckon the robot will not be useful in most disciplines.
Meine Schwester ist sehr, sehr reizvoll - Nietzsche
Science requires some kind of passion/imagination/interest to start. After that, you employ scientific method to create knowledge. But, I don't think we fully understand the first part.
Computers are useless. They can only give you answers.
-- Pablo Picasso
"In a number of areas scientific data is being generated at enormous rates, creating the need for the automated analysis of the data," said Ross D. King,
So basically it collects hundreds of terabytes of data, then uses certain algorithms to analyze it in an effort to try to spot a trend.
So far so good, but the part where it tries to interpret the data in a more innovative way by creating theories is for me the breakthrough. I can't help but think that credit (if a new theory is discovered) must go to those who wrote the algorithm.
My question is: If it invents something otherwise patentable, who files the patent -- and would such a patent be enforcable?
Free Software: Like love, it grows best when given away.
And then it is just a short step using this to stop scientific research unless they get a cut, because it would be unauthorized use of their patented processes and methods. Even if implemented in a biological system like a brain
"It is a greater offense to steal men's labor, than their clothes"
That's just the first step in the analysis. The fun starts with: 'does the person who invented it have the right to patent it's inventions? Does the person who owns it have the right (if different than the person who created it)?'
Free Software: Like love, it grows best when given away.
Do a google for "lenat" and "eurisko" and you'll find a system that did this thirty years ago. Designed by Doug Lenat, Eurisko was a software that created and tested new mathematical theorems. Didn't evolve much after that, because there's a lot more to science than just creating and proving theorems.
While I'm sure this was an attempt at humor, Asimov's Laws of Robotics aren't practical in terms of our current understanding of AI.
Any system complicated enough for such laws to have meaning would also be complicated enough that it couldn't be created by explicitly programming the intelligence. It would have to be a system programmed by teaching, and it would be difficult (if not impossible) to give it rules that were absolute overrides, yet high-level enough to require much of what they've learned just to interpret.
It is a nice joke, but it is a scary truth in it. Once man(kind) will invent an artificial inteligence which is superior to him, the "critical mass" of the "evolution of intelligence" is crossed and chain reaction - fast exponential grow of that inteligence can start. I have no doubd that this crossing is possible, the question is when (in 30, 100 years?).
PS. The idea is not mine, Stanislaw Lem wrote about it