English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

Deep neural networks ("deep learning") have emerged as a technology of choiceto tackle problems in natural language processing, computer vision, speechrecognition and gameplay, and in just a few years has led to superhuman levelperformance and ushered in a new wave of "AI." Buoyed by these successes,researchers in the physical sciences have made steady progress in incorporatingdeep learning into their respective domains. However, such adoption bringssubstantial challenges that need to be recognized and confronted. Here, wediscuss both opportunities and roadblocks to implementation of deep learningwithin materials science, focusing on the relationship between correlativenature of machine learning and causal hypothesis driven nature of physicalsciences. We argue that deep learning and AI are now well positioned torevolutionize fields where causal links are known, as is the case forapplications in theory. When confounding factors are frozen or change onlyweakly, this leaves open the pathway for effective deep learning solutions inexperimental domains. Similarly, these methods offer a pathway towardsunderstanding the physics of real-world systems, either via deriving reducedrepresentations, deducing algorithmic complexity, or recovering generativephysical models. However, extending deep learning and "AI" for models withunclear causal relationship can produce misleading and potentially incorrectresults. Here, we argue the broad adoption of Bayesian methods incorporatingprior knowledge, development of DL solutions with incorporated physicalconstraints, and ultimately adoption of causal models, offers a path forwardfor fundamental and applied research. Most notably, while these advances canchange the way science is carried out in ways we cannot imagine, machinelearning is not going to substitute science any time soon.

Off-the-shelf deep learning is not enough: parsimony, Bayes and causality