Pages

Tuesday, August 2, 2011

What if computers could think?

 
Machines have overcome humans in many tasks involving simple logic and strength, but in the matter of ambiguous situations or complex problems none of them can be compared with human intelligence. It would be a great advance in technology if computers systems could learn by themselves and improve their previous knowledge by experiencing new information (like we do!). To make this picture clear, I’ll show you a basic example:

Imagine that you, every Friday night, go down to the same bar and request a pineapple and mint juice. As weeks go by, the bartender will already know what you’ll order and make it for you automatically (but that doesn’t change the fact that he stills considers you a nerd who drinks juice on a Friday night instead of ordering a booze). That happens because he learned through experience the same event over and over and noticed a pattern!

So, in order that computers be able to learn, they must identify patterns. There are many programming techniques and running systems you use them all the time, such as your web browsers (like Explorer, Chrome, Mozilla firefox), if you type the letter “F” inside the address bar, probably it will show up “facebook.com”, that happened because most times you type “f”, you ended up writing “facebook.com”. For a bunch of wires, boards and code lines it’s pretty smart, isn’t it?

 Of course this is just an ordinary and commercial application. It will cause a dramatic change in society if these systems learn from medical records, for example, and discover new kinds of treatments or vaccines for AIDS, cancer and other diseases. In the economics field, they could predict what kinds of investments are the most profitable and secure. Sounds amazing, doesn't it? But unfortunately these solutions belong to the future, nowadays we can’t obtain a relevant performance yet, such as a human being, but we are not too far from it.


An attempt of implementing a machine learning system is NELL (or Never-Ending Language Learner) which is a Natural Language Processor, this system simply reads the web continuously and learns from the huge amount of information present on it. NELL started as a project inside Carnegie Mellon University (USA) and 2 years ago the system run for the first time. Professor Estevam Hruschka Jr, who is one of the researchers of this project, is also a faculty member in UFSCar and for sure the Brazilian authority in Machine Learning. For the past 6 months I had the pleasure to work with him in a brief university project that allowed me to acquire plenty of knowledge and renewed my passion for computing, it was a nice experience that I hope to explore even more.


You can check this video to see Tom Mitchell (world’s authority in machine learning) being interviewed, it’s very informative, take a look!
http://www.techcrunch.tv/watch?id=1yZTlyMTqnbxc0F_wYjkcMt4tOofe5Q1#ooid=1yZTlyMTqnbxc0F_wYjkcMt4tOofe5Q1

Here’s NELL website, you can check all the things he learned. Last time I checked he didn’t know what Barone English is, but he did know that Mary it’s the mother of Christ, and that is correct!


Mary is the mother of Christ, and NELL is 100% sure!
Among the great advances thinking computers could bring to us, deep down in all of us lies fear, yes dear reader: FEAR. Once I was telling a story to my uncle about computer’s intelligence, I was just saying that in the future computers will become so smart that they could beat chess winner champions easily or perform simple human activities and how cool it could be and that’s when he just snapped! He started yelling that it’s not possible a thing like that being realistic, and if it were possible to achieve, it shouldn’t be allowed by the government because computers would take our jobs and generate unemployment, economic chaos, and afterwards when machines get smart enough, they’ll dominate us completely as they don’t have feelings, like a Tech-Dictatorship the movie Terminator described.

Vocabulary:
to make this picture clear: to clarify things.
booze:  any kind of alcoholic drink.
bunch of: a lot of.
to snap:   become extremely angry or upset because a situation has become too annoying or difficult.
mean (informal adjective) cruel, or unkind.


Scene from the movie 
Terminator 3: Raise of the Machines 

Can computers become mean?
Some people may agree with my uncle in some degree, but others (like me) find those statements hilarious! But when he said “machines don’t have feelings” it’s perfectly correct, after all they can’t feel anything, at least not any human feeling like compassion, gratitude, hate or love. His point of view doesn’t even make sense and it only would be true if your cell phone for instance mixed up or deleted all your contacts just to make you frustrated. Machines only do stuff that are programmed for them to do, so in the future we won’t be at war against mean robots, unless someone programs them to be mean…
I hope that I could separate science fiction from real science in this post and you will not afraid to use your computer or any machine next time! ;D
This was my first post in the blog, I really enjoyed the experience and I’ll gladly write again if you all liked it! Thanks for the opportunity Vivian!

See ya!


by Arnon Rambaldo
undergraduate in
Computer Science at UFSCcar

No comments:

Post a Comment