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."
From the article:
The system, say its British creators, did just as well as biology graduate students in solving a problem in genetics, according to an article in today's issue of the journal Nature.
In other news, a calculator does just as well as a PhD mathematician at solving arithmetic problems.
Come on, it's a neat invention, but it's solving a closed problem-- not worthy of being called a scientist.
Chance favors the prepared mind. -- Louis Pasteur
If not, it won't do well, besides the lack of ability to think creatively.
-Cyc
/.'s 10 Millionth
In the late 1800s mathamatitions had this idea that you could write a bunch of rules that would allow undergraduates to devise proofs. This had a lot of interest until Godel (and others) proved that it can't be done.
In traditional /. fashion I didn't read the artical. Still it seems to me that either this is very limited in what it can research, or it can't work. If it is limited, there isn't much news about a robot programed to do something either too repeatative for a human to finish, or too dangerious for a human to do. If it can't work, well I still welcome the limited expiriments it can do which can enhance knowledge, if we don't treat it like the end of all science when this machine does all it can do.
all of human invention is about laziness. Try'n think of a single invention that doesn't make a task easier or quicker. This isn't to say that science is about laziness, science is about the persuit of knowledge. Engineering is about laziness, though.
write proposals? That seems to be 95% of my advisor's job. No science can be done without the money to do it with.
When will this robot find some interesting theory and experimental proof that qualifies it for a Nobel prize? (Or would qualify it for the prize if a human had done the same work?)
This invention demonstrates the full power of computers to mass-produce logical human thought processes. Although it may be very hard to reduce the mental processes behind creating theories and experiments to a set of algorithmic processes, once done the possibilities are endless. A robotic scientist can be mass produced for far less money and in far less time than it takes to grow a new Ph.D person.
Software is, in my opinion, a more powerful invention than was writing. While writing encodes and distributes static thoughts, software encodes and distributes the dynamic thought processes.
Two wrongs don't make a right, but three lefts do.
I don't know, maybe it won't invent the cure for cancer, maybe it won't be able to decode the sequence and meanings of life - but just like a calculator, it will automate known procedures. This will, at the very least, increase the efficiency of what human scientists can do. I agree, it is limited to what it has been programmed to do. The AI portion is probably not advanced enough to figure out extremly complex, unknown issues (and it probably doesn't get things like 'hunches.'). But considering that figuring out how yeast cells work is a lot more complex then a calculator, it is still an impressive piece of technology, that will (hopefully) help scientists out. -Avi
I mod down so you can mod up. Your welcome.
Just wait 50 years or so. Once we get sentient computers you can bet there's gonna be a class-action filed on behalf of all these creative boxen.
Then the courts fail to recognize the boxen as entities, the war starts, and we're in one of about a half-dozen terrible movie universes.
I wonder if the computers will kill the smart reasonable humans too. I suppose I should be keeping all these old Linux CDs to present as evidence at my trial. . . . . . .
Only in a Slashdot fantasy can a Slackware install turn into several hours of sex . . . . .
This could be wonderful news for the advancement of science in general. Most of it is trial and error. Mix these 2 together and see if has the desired effect. There may be 10 or 20 thousand combinations to try. That's what experimental science is all about. Now if a grad student could just setup one of these things to test all combinations until either the wanted result appears, interesting things not predicted happen, or favorable or disfavorable results happen that could be useful else where. I could see a robot testing combinations until a given event is true. How would software flag "interesting results?"
Example: Scientist is looking for non-stick film to apply to pots. Robot is testing combinations. Does it notify the scientist if say this combination makes the pot super conductive, but things still stick to it?
Just like the "programmer creates programmer" thing. Code generators have been around for ages, but I have yet to see a program that can think up a program for itself (or even turn a requirements-document into actual working software).
Lots of people are commenting that this isn't that useful because the robot won't come up with new scientific breakthroughs. But I suspect that none of these people have actually done biotech lab work.
Lab work largely consists of doing the same thing over and over and over and over. My partner is doing a PhD in molecular biology, and I have spent more than a few nights and weekends helping her by being a robot. For example, one Sunday I spent about 10 hours gathering "growth curve" data. This involves taking dozens of vials of growing yeast, and measuring their optical density every 2 hours or so. To do this, you take the vials out of a spinning wheel, put them in a tube holder, carry the tubes to a desk, put new tips on a pipette, mix the tubes to stir them up again, suck out some of the fluid, and squirt the fluid into a smaller tube. Then you put the large tubes back, carry the little tubes to the optical density device, insert them, run the measurement, print out the results, pull out the little tubes, put them in a styrofoam holder for posterity, and repeat.
This process was incredibly labor intensive -- I had about 10 minutes of rest time every 2 hours, over the course of 10 hours. And after those 10 hours my partner took over and continued the process for another 10 hours.
Not only would a robot have been a welcome relief to this process, we actually spent quite a while discussing the specific requirements and possible design of such a robot.
A robot like this is useful because it provides the equivalent of a compiler and automated test suite. The interesting things in biological science do not come from grad students running through the grunt work manually -- they come from grad students using their brains to design the experiment and then analyze the results.
Obviously this robot won't replace the grad students entirely. But it might let them be vastly more productive.