I side with those who feel they must agree with Elon Musk that Yann LeCun is full of shit when he claims that nothing that is not published counts as “science”.
Among the respondents to a Quora answer that I agree with, one defender of LeCun says “In the broad sense, science is a body of knowledge that relies on repeatability and acceptance, which is hard to accomplish if the work is not known. A better statement might be if it’s not published, it’s not adding to science.” But I don’t agree with that definition of science.
One problem here is with the word “acceptance” and the question of the required scale (and nature) of the accepting community. Few would deny that an alien civilization might have a practice that we would call science despite the lack of publications in Earthly journals, and by the same logic it is hard to deny the label of “science” to a similar practice restricted to an audience that is small enough for word of mouth to be the only necessary means of communication.
Another commenter suggests broadening the interpretation of “publish”, saying that scientific work may be “internally” published within a small group, but adding that “Nonetheless, your work would still be open for critical examination, which is the key”. But I’m afraid I don’t even agree with this, as I can imagine a person doing perfectly valid scientific work in complete secrecy. It’s not that the work has to be open for critical examination, but that if it were, then it would turn out to be accepted and would lead to teachable methods for making accurate predictions.
Of course anyone who wants to is free to use the word “science” only for the body of work that is globally accessible in some list of specific refereed journals, but (as with most words) there is no universal consensus as to what is the “correct” usage.
Source: (1001) Christopher VanLang’s answer to Does science need to be published? – Quora