In natural language processing, currently, research is published on arXiv and reviewed on Twitter. This is not how the institutions and funders involved would d

In natural language processing, currently, research is published on arXiv and reviewed on Twitter. This is not how the institutions and funders involved would describe it, but it's what actually happens. If you're waiting for it to show up in a peer-reviewed journal, you're months or years behind. Conferences are a good way to get caught up by letting the discussion of a number of papers happen in one physical place. The publishing industry is not necessary to organize conferences. Journals seem to be a formalized mechanism for keeping score of the number of positive, novel results you've produced, in case you're ever up for tenure. Keeping score that way is not great for science anyway, because it biases science against negative results and replications. When papers are discussed in the open on the Internet, you can see the discussion and choose who to believe. In formal peer review, you get the opinions of a few anonymous, uncompensated people, who would rather be doing something else, aggregated into a single bit.