Other branches of physics have definitely been more responsive to the latest developments in machine learning. Leaving aside futuristic arguments about when, if ever, robotic systems will replace scientists (Hall, 2013), we think this is an excellent time to think about AI for a Space Weather scientist, and to try formulating (hopefully realistic) expectations on what our community can learn from embracing AI in a more systematic way. The range of applications is indeed very vast: fraud detection (Aleskerov et al., 1997), online product recommendation (Pazzani & Billsus, 2007 Ye et al., 2009), speech recognition (Hinton et al., 2012), language translation (Cho et al., 2014), image recognition (Krizhevsky et al., 2012), journey planning (Vanajakshi & Rilett, 2007), and many others. In fact, one might not realize that, for instance, most of the time we use an app on our smartphone, we are using a machine learning algorithm. Still, many reckon that this time is different, for the very simple reason that AI has finally entered industrial production, with several of our everyday technologies being powered by AI algorithms. This might or might not be followed by yet another winter. Even the initial Dartmouth workshop held in 1956, credited with the invention of AI, had underestimated the difficulty of understanding language processing.Īt the time of writing some experts believe that we are experiencing a new AI spring (e.g., Bughin & Hazan, 2017 Olhede & Wolfe, 2018), which possibly started as early as 2010. Indeed, it is unfortunate that most of the AI research of the past has been plagued by overconfidence and that many hyperbolic statements about utility of AI had very little scientific basis. Such a cyclical trend is not atypical for a potentially disruptive technology, and it is very instructive to try to learn lessons from (in)famous AI predictions of the past (Armstrong et al., 2014), especially now that the debate about the danger of artificial general intelligence (i.e., AI pushed to the level of human ability) is in full swing (Russell & Bohannon, 2015 Russell & Norvig, 2016). The history of artificial intelligence (AI) has been characterized by an almost cyclical repetition of springs and winters: periods of high, often unjustified, expectations, large investments, and hype in the media, followed by times of disillusionment, pessimism, and cutback in funding. The clearest opportunity lies in creating space weather forecasting models that can respond in real time and that are built on both physics predictions and on observed data.ġ Artificial Intelligence: Is This Time for Real? If the community can master the relevant technical skills, they should be able to appreciate what is possible within a few years time and what is possible within a decade. The challenge is that the current data science revolution has not been fully embraced, possibly because space physicists remain skeptical of the gains achievable with machine learning. Today, machine learning poses both a challenge and an opportunity for the space weather community. Space weather is a discipline that lives between academia and industry, given the relevant physical effects on satellites and power grids in a variety of applications, and the field therefore stands to benefit from the advances made in industrial applications. In particular, the combination of massive data sets and computing with specialized processors (graphics processing units, or GPUs) can perform as well or better than humans in tasks like image classification and game playing. In the last decade, machine learning has achieved unforeseen results in industrial applications. The recurring themes throughout the review are the need to shift our forecasting paradigm to a probabilistic approach focused on the reliable assessment of uncertainties, and the combination of physics-based and machine learning approaches, known as gray box. On the other hand, this paper serves as a gentle introduction to the field of machine learning tailored to the Space Weather community and as a pointer to a number of open challenges that we believe the community should undertake in the next decade. On one hand, we will discuss previous works that use machine learning for Space Weather forecasting, focusing in particular on the few areas that have seen most activity: the forecasting of geomagnetic indices, of relativistic electrons at geosynchronous orbits, of solar flares occurrence, of coronal mass ejection propagation time, and of solar wind speed. This Grand Challenge review paper is focused on the present and future role of machine learning in Space Weather. The numerous recent breakthroughs in machine learning make imperative to carefully ponder how the scientific community can benefit from a technology that, although not necessarily new, is today living its golden age.
0 Comments
Leave a Reply. |
AuthorWrite something about yourself. No need to be fancy, just an overview. ArchivesCategories |