We are happy to welcome Manuel Weber, Particle Physicist and Postdoctoral Fellow at Stanford University, as a guest on our blog.
With the rise of artificial intelligence, or AI, our world is changing and it may seem to come straight out of a science fiction movie what these systems can do already. At the same time, the technology is still very immature and we’re only starting to explore its full potential. The ‘intelligence’ in AI should probably be considered an overstatement as we’re still far from building truly intelligent systems. Currently, what we call AI is more properly described as deep learning and an attempt to mimic the functionality of the brain with self learning computer algorithms that consist of millions of neurons, hence the name artificial neural network. The success of these algorithms is largely due to the shear amount of data we produce every day. Every picture we upload to Facebook or Google, every conversation we have with Siri and every click within that online store we like so much, helps these networks learn from more data and therefore become smarter. And it gets event better: Today AI makes cars drive by themselves or help doctors diagnose diseases.
Nothing new for particle physicists
The immense potential of AI has led the big companies to invest a lot of money in further developing the technology and make it accessible to a broader audience. It is for this very reason these techniques are now making their way into fundamental sciences. For researchers in particle physics, like myself, this is nothing new though. The idea of using artificial neural network to disentangle specific detector signals was already applied 30 years ago and more recent discoveries, including the Higgs particle, made extensive use of AI. Advancements in computer vision, through the use of neural networks, have led particle physicists to rethink their approach to analyzing the huge amount and complex data. To put it in perspective: Experiments at the Large Hadron Collider in Geneva produce more data in an hour than Facebook in a year. A recent challenge on Google owned platform Kaggle asked AI experts around the world to tackle this problem and granted $25’000 for the three best teams. Similar challenges in the past have produces some very impressive improvements in AI.
There is an extensive potential for AI to become a standard tool for scientists in many different areas. At the cutting edge of scientific discoveries innovative approaches are necessary. There is an increasing amount of publications, which demonstrate the use of AI in their research. My own research team at Stanford University has just shown how AI provides a legitimate alternative to classical analysis methods in our latest publication. And the results are remarkable. Other particle physicists are on the same quest as a recent article in the Nature magazine summarizes.
We are in the middle of technological revolution and AI experts are not shy to call it the new electricity. As with every new technology, we need to learn how to live with it and use it to our advantage. As for science: I hope we’ll see some exciting new advancements by incorporating AI into our research and make many new discoveries.