The journal Nature Machine Intelligence was launched in January this year and is already setting new standards for transparency and reproducibility of research findings in the field of machine intelligence. Together with Nature Methods and Nature Biotechnology, Nature Machine Intelligence launched a trial in partnership with Code Ocean which enables authors to share fully-functional and executable code accompanying their articles and making it easier to review the code and replicate the results.
“The uptake from authors is very high. It's been interesting and very heartening to see that authors are really willing to work with us on sharing data and code and are enthusiastic about taking part in the Code ocean trial,” says Liesbeth Venema, the journal’s chief editor.
She also introduced a new type of article called Challenge Accepted. It is a one-page article from scientists who have organized, won or taken part in a particular competition in data science, robotics or machine learning. According to Venema, competitions are important because they can stimulate new directions for a field, help to concentrate on overcoming a particular hurdle and offer another way for scientists to share resources with each other. “They also play a big role in giving young scientists a chance to show what they can do, learn new skills and to work as a team,” she adds.
Highlights so far have been a research paper about a limb that can teach itself to walk as well as a new AI approach to solve a Rubik's Cube that can also be used to solve much bigger combinatorial puzzles which is still a big challenge in the field.