Can a computer discover scientific laws?
In his first column for Seed magazine, my Institute for the Future colleague and pal Alex Pang looks at efforts to create software that doesn't just support scientific discovery, it actually does new science. From Seed:
Older AI projects in scientific discovery tried to model the way scientists think. This approach doesn’t try to imitate an individual scientist’s cognitive processes — you don’t need intuition when you have processor cycles to burn — but it bears an interesting similarity to the way scientific communities work. (Cornell professor Hod) Lipson says it figures out what to look at next “based on disagreement between models, just as a scientist will design an experiment that tests predictions made by competing theories.”Why We're Not Obsolete: Alex Pang in Seed
But that doesn’t mean it will replace scientists. (Cornell graduate student Michael) Schmidt views it as a tool to see what they can’t: “Something that is not obvious to a human might be obvious to a computer,” he speculates. A program, says Schmidt, may find things “that look really strange and foreign” to a scientist. More fundamentally, the Cornell program can analyze data, build models, and even guess which theories are more powerful, but it can’t explain what its theories mean — and new theories often force scientists to rethink and refine basic assumptions. “E=mc2 looks very simple, but it actually encapsulates a lot of knowledge,” Lipson says. “It overturned a lot of older preconceptions about energy and the speed of light.” Even as computers get better at formulating theories, “you need humans to give meaning to what the system finds.”
the latest
latest episodes
I'm doing science and I'm still alive
In my work on the narration of "Edison His Life and Inventions", I learned what an exhaustive experimenter he was. The Edison battery was the end product of 50K of experiments. While a computer is a handy modeling tool it won't tell you that some initial assumptions or theory are faulty or plain wrong. There are some answers that can be approximated by computers but never solved. It takes some intuition to work around dead ends such as that.
This is old news. There was a special on the news about 10 years ago about a computer that discovered Snell's law from refraction data.
50K trials? Am I to conclude that Edison didn't have any intuition? It sounds like a modern computer scientist could code a replacement for him.
Wolfram Alpha anyone?
The article seems to ddescribe some sophisticated data mining. It makes the common mistake of confusing laws with theories. Laws are useful predictive tools that don't necessarily explain why anything happens. For example the "Law of Gravity" decribes the mathematical, quantitative relationship between mass, distance and force. It is very useful for making spacehips go where you want, but it says NOTHING about why gravity exists at all.
A theory, on the other hand, tends to be more, well, theoretical. It goes to explanations of things. A much harder proposition for a computer.
Could computers come up with laws? Yep, probably. Could they come up with theories? Not likely IMHO.
I love anybody who quotes Harry Collins.
Can a computer discover scientific laws?
Absolutely no.
Science is empirical, meaning that knowledge arises from experience, and evidence. Unless the computer can set up experiments, a computer cannot discover scientific laws.
That is why computer models are not science, but cargo cult.
Well, you can see computers doing experiments on other planets, then sending the data here. If it would crunch the data there and find a correlation, I guess it would have discovered something.
Now, someone had to predict what kind of data and correlation there would be. Today, there's always someone behind everything a computer do. When robots will be able to be fully aware of their environment and implement that Cornell program, it could look like a " sciencist ", since they say the program is able to start by himself different tests. Sure, it's still a far stretch, but it looks possible, in a way, and finding mathematical relations isn't everything as people already pointed out. It would be a lot more fun for the scientists to concentrate on the qualitative part of the science, and trying to make sense of all the different relations that would pop up everywhere (that part would maybe call for a meta-robo-scientist?).
I'm a physics student, and I don't feel threatened at all by all this.
Also, WinkyBB, I don't think anyone can say "why" gravity exists. The simplest answer to that is that we see something consistent in nature, and we measured it. That's why we have a model we call "gravitation".
Yes or now,
depending on how you define 'computer', 'discover' and 'scientific laws'.
If you define 'computer' as 'the beige box on my desk' and 'discover' as 'doing something uniquely human', then of course the answer is no.
If you define 'computer' as 'a programmable machine' and 'discover' as 'find and test new information', then yes. Computers are doing so right now. In fact, my computer is doing so right now.
It's a question of definitions, and debating them will, I fear, not shed much light on neither computers nor science.
Au contraire, my dear Spazzm, while you are hashing out those fine distinctions there are many things you might learn; not the least of which is what your own opinion really looks like, examined up close.
Too many polite, well-meaning people shy away from debating definitions, a deplorable trend...
ever heard Gödel?
I once started a book about something he theorized, but I never completed it.