Computer Vision News - November 2018

intellectual property. The whole point of this is the exchange of information between industry and academia so that you can take advantage of the fact that in industry you have engineering support and large computing facilities, and in universities there are students and young people with a lot of creativity whose incentives and modus operandi is different from industry. It’s good to have different motivations for people because they come up with different ideas when they’re in different environments. That’s good, but it only works if the industry lab you are in practices open research and actually does real research with publications. It doesn’t work if the affiliation is with an industry where everything is secret, everything is very applied and engineering-oriented. There’s been some responses to my piece and some of them basically blurred the difference between these two kinds of industry research. They said you can’t have dual affiliation, because when you work in industry you have to basically work on whatever is useful to the company, and keep some things secrets, and that’s incompatible with academia. I agree, it’s incompatible, but I disagree that it’s impossible to do this in industry. It depends how the industry research is organised. One of the things I’ve done in Facebook is organise the research lab in such a way that it is compatible with academic collaborations. That’s probably my big contribution to industry research within the last five years. It’s something that basically did not exist to the same extent until now. There have been organizations in industry in the past that have been very influential in science, like Bell Labs where I used to work and IBM Research and Microsoft Research, 10 Computer Vision News “That’s probably my big contribution to industry research within the last five years…” Yann LeCun Guest