The larger credit for the deep learning revolution has always been attributed to faster and larger computers, and also to the availability of larger datasets. But if you ask researchers and developers in the field of machine learning, they will point to the real hero of the revolution: open access. The fast and open discovery and dissemination of state-of-the-art knowledge is truly fabulous.
But this movement has had a small distraction with a recent development in the artificial intelligence ecosystem.
Recently, Springer Nature, which is the publisher of Scientific American announced that it will launch a new journal in 2019, titled Nature Machine Intelligence. This is seen as a move to encash into the sky-rocketing field of ML and AI. But one of the problems with this great new endeavour was the fact that this new journal is not openly accessible. All the research done for and submitted to the journal will be put behind a paywall.
Boycotting Paid Machine Intelligence Journals
Thomas Dietterich, former editor of the Journal Of Machine Learning made a statement saying, “…Journals should principally serve the needs of the intellectual community, in particular by providing the immediate and universal access to journal articles that modern technology supports, and doing so at a cost that excludes no one.”
The statement is supported by more than 3,000 AI researchers including many legendary names. Their view is that there is no role for closed access in the artificial intelligence research and the new closed-access journal is a step back. The researchers came out in support of “zero-cost open access journals and conferences in artificial intelligence and machine learning.” This model is similar to arXiv. The open debate of corporate takeover of research outputs that are financed by taxpayers money is slowly taking centre stage. The work done by the researchers is paid by the university and indirectly by the taxpayers.
Machine learning researchers have historically been adopting the open source and open access ecosystem. There are excellent repositories of knowledge such as the peer-reviewed Journal of Machine Learning Research and arXiv, among others, who support open access. This is not only limited to research output but also encompasses the machine learning tools. This can be seen in Google’s TensorFlow, PyTorch, Microsoft’s Cognitive Toolkit and others. The world is having a great amount of open source code that can be put to great use. But according to Gartner’s estimates only 15 percent of companies or organisations get into production with ML or AI.
Advantages Of Open Access And Open Source In ML
Google open sourced Tensorflow stating that, “We hope this will let the ML community – everyone from academic researchers, to engineers, to hobbyists – exchange ideas much more quickly, through working code rather than just research papers.” There is obviously incentive in giving out free tools and knowledge. The publication of research on arXiv has led to a rapid development of ML algorithms which in turn have led state-of-the-art results in natural language processing, computer vision and other fields.
The open sourcing of tools and knowledge is a win-win situation. The knowledge produced does not remain in the hands of the few. It is widely shared and students, researchers from even the most remote parts of the world can pace up to the forefront on AI via these free tools and access to state of the art models. And on the other hand With TensorFlow, Google is looking to bringing ML and AI to a large section of the world, and then further encourages engineers to run their projects on their Google Cloud. It is one of the most meticulous long-arching strategies that is turning out to be doing good for all sides.
Old Conflicts In The New Era
Going back to Springer Nature’s Journal story, the irony is that even Nature Machine Intelligence took to arXiv to help them monetise their closed access model. A representative for Nature Machine Intelligence said that the journal encourages open access ideas and it does this by letting authors in the journal to put up the preprint versions of their papers on platforms such as arXiv and SharedIt. But still the paper on the Nature Machine Intelligence platform remain open only to paying academics and corporates who can afford to pay hefty access fees.
The battle like this one, with corporate publishers on one side, and researchers and public on the other side is nothing new. But this conflict has appeared at a point in time when the AI revolution is on full throttle. Such issues should be sorted out sooner rather than later, otherwise it has the power to put the revolution in danger.
Try deep learning using MATLAB