Speeding up prevention and treatment? That’s the hope of Kartik Hosanagar, a Wharton University artificial intelligence analyst. Introducing his new book, A Human’s Guide to Machine Intelligence: How Algorithms Are Shaping Our Lives and How We Can Stay in Control, he offers AlphaGo as the way to Go:
An alternative approach is illustrated by Google Deep Mind’s Go-playing software AlphaGo Zero. While the original version of the software was trained heavily on past games played by human Go players, AlphaGo Zero didn’t bother studying human moves; instead, its entire training dataset was self-generated. The software, armed with basic rules of Go, played millions of games against itself. Next, it analyzed those games to figure out which moves helped and which ones hurt.Kartik Hosanagar, “Can Artificial Intelligence Make Scientific Discoveries?” at The Scientist
Analyst Eric Holloway isn’t quite so sure. He told Mind Matters News,
I agree there is some extent that automating physical experimentation is useful.
However, it has limited usefulness. The idea is not new, it is well tested in the realm of machine learning. Machine learning essentially takes statistical hypothesis testing from the sciences and applies it to finding models that fit the data.
The rub lies in “fitting the data.” If the model is complex enough then it is easy to “overfit” the data, wherein the model has not learned anything but the data itself. This is also called “memorizing” the data.
On the other hand, if the model is too simple, then it will never be able to learn anything, regardless of the amount of data used.
What is needed is the happy medium of a model that is complex enough to learn, but simple enough to not memorize. There is no way to automate finding the happy medium. It requires human intelligence.
It appears that the Machine that will do our thinking for us and automatically get everything right isn’t physically possible. We can use the machines but we will still have to think for ourselves.
Also by Eric Holloway: Artificial intelligence must be possible! Really? Many arguments for strong artificial intelligence depend on an ideological commitment to explicit, unproven theories about the universe
Human intelligence as a Halting Oracle