Publishing, Peer Review and Preprints: Numenta's Strategy for Sharing Science
Occasionally we receive questions about our publishing strategy, such as why we share papers that have not yet been published in a scientific journal, or what is our stance on open science. Our co-founder, Jeff Hawkins, recently responded to a question about this topic that was posted on our HTM Forum. For those of you who are not active there, you can read a consolidated version of Jeff’s reply below.
Jeff Hawkins, Numenta Co-Founder
Peer-reviewed papers have been the standard and preferred form of publication for many years. They are essential for career advancement in academia, and academic scientists are often under pressure to publish many papers. Sometimes this leads to papers that are of questionable value, but overall the system works well.
Industrial research labs have less of this pressure and it is common for them to publish non-peer reviewed research notes on their websites. Many people appreciate being able to access these even though they are not peer-reviewed. Major advances from industrial labs will typically be published in journals, but not always. Sometimes they are considered trade secrets and the scientists get upset if they cannot publish their results.
It is a lot of work and takes a long time to get a peer-reviewed paper published. After submission, it may take a few months to a year. This is such a long time that scientists have started posting their manuscripts to preprint servers while waiting for the paper to be accepted by a journal. This began in physics but has now spread to other fields. The big preprint server for neuroscience is “bioRxiv”. Most journals permit this, but not all. The final paper, when approved, will be similar to what is on the preprint server. Often a lab or university will put out a press release when the preprint version is available and not when the final paper is accepted. It is generally considered ok to cite preprint articles as long as you change your citation when and if the final paper is accepted. There is a movement underway to make preprint servers the final publishing forum and open them to peer review in a wiki-like fashion.
There are two types of journals. Open access and closed access. This is a huge topic (easy to find articles about it). To many people (me included) the old model of closed access is broken as it can be prohibitively expensive to read articles on closed journals. Unfortunately, some of the most “prestigious” journals are closed access. This is changing as more and more scientists refuse to publish in closed journals.
Ok, what about Numenta, what do we do to publish our results? It takes many months to write a paper and many more to get it published, so in the past we chose to work more on the science and delay publication. Our peer-reviewed paper, “Why Neurons Have Thousands of Synapses, A Theory of Sequence Memory in Neocortex” was published in 2016, five years after we posted a whitepaper on our website describing the same algorithm. We now have five journal papers total and three more being written. This is a pretty good rate given the size of our team. (Some scientists may work on one paper for several years.) Most journals keep track of statistics such as how many times a paper has been downloaded and how many times it has been cited. If you are interested you can track the statistics for our papers.
Peer-reviewed papers are not the only means we use to publish. We also present posters at conferences. These can be very competitive and undergo peer-review to get accepted. We post all our conference posters on our website. We also give invited talks and, when we can, we post links to the talks. Finally, we post smaller papers and research notes exclusively on our website and on preprint servers. Usually we do this because the topic has a limited audience or represents a sub-topic of something we published elsewhere.
Our goal is to make all of our research accessible. We submit our major discoveries to be published in open access peer-reviewed journals, while other work will be shared on our website. Ultimately the impact of our work will not be determined by the number of papers we publish or where they are published, but by whether the ideas we are bringing to the table are significant and correct. I am confident they are, but time will tell.