3 ms·

You are saying that the risks of industry interests have not come true. There has been much talk about companies trying to influence what products are being use

by cjfd 5y ago

You are saying that the risks of industry interests have not come true. There has been much talk about companies trying to influence what products are being used in education, which for universities, is closely related to research. Even if we assume this to be true a risk that has not happened (much) is by no means guaranteed to not happen in the near future. Also the examples that you name "bandwidth compression, codecs, recommender systems, crytocurrency" are really quite sexy. They sound like things academics would research all on their own without corporate involvement. The thing is that basically these things can be researched if one just has a computer, or in the case of bandwidth compression, a few computers, and enormous amounts of money are not really necessary so the corporate involvement would not seem to be strictly necessary. And then, when a corporation is involved one runs the risk of the findings disappearing behind a copyright or a patent wall. It is much better for all of us if they become available to everybody. That is, by the way, one of the reasons to have academia in the first place.

srvmshr 5y ago
> There has been much talk about companies trying to influence what products are being used in education, which for universities, is closely related to research. I will speak for ML & Systems. I am not an expert on all domains. Most tools that researchers use in ML are open-sourced e.g. Pytorch TF Keras etc. Foundational papers like GFS, BigTable, Hadoop etc are publicly available now. There is a moratorium on how soon it appears, but it isn't behind walls forever. Academia tends to choose more open source over closed source. I'd have argued MATLAB to be more successful than Python in that case. It is not. In industry the practice may be more of 50-50 or even more towards closed source. > They sound like things academics would research all on their own without corporate involvement. The thing is that basically these things can be researched if one just has a computer. How can you emulate operation involving scale with just a computer or a bunch of computers. Hence, that is where industrial efforts come in play. A lot of things work well for a dozen, and then a completely different problem emerges when we talk about hundreds or thousands of users. I am only alluding to the fact that CS, in contrast to many other disciplines, has a more symbiotic relationship with industry. Companies do have incentives of using copyright & trade secrets, but there is enough trickle-down effect that gives academia to pursue newer challenges. The cycle repeats over and over. Academia cannot replace industry and likewise. If there is any pressing problem of this symbiosis, it is more of labor attrition. More people are leaving academia for better pay. But that is not what the topic was about.
nyanpasu64 5y ago
> And then, when a corporation is involved one runs the risk of the findings disappearing behind a copyright or a patent wall. True, but universities often patent algorithmic (and other) discoveries and try to charge companies for implementing them as well.

by cjfd 5y ago

srvmshr 5y ago

> There has been much talk about companies trying to influence what products are being used in education, which for universities, is closely related to research. I will speak for ML & Systems. I am not an expert on all domains. Most tools that researchers use in ML are open-sourced e.g. Pytorch TF Keras etc. Foundational papers like GFS, BigTable, Hadoop etc are publicly available now. There is a moratorium on how soon it appears, but it isn't behind walls forever. Academia tends to choose more open source over closed source. I'd have argued MATLAB to be more successful than Python in that case. It is not. In industry the practice may be more of 50-50 or even more towards closed source. > They sound like things academics would research all on their own without corporate involvement. The thing is that basically these things can be researched if one just has a computer. How can you emulate operation involving scale with just a computer or a bunch of computers. Hence, that is where industrial efforts come in play. A lot of things work well for a dozen, and then a completely different problem emerges when we talk about hundreds or thousands of users. I am only alluding to the fact that CS, in contrast to many other disciplines, has a more symbiotic relationship with industry. Companies do have incentives of using copyright & trade secrets, but there is enough trickle-down effect that gives academia to pursue newer challenges. The cycle repeats over and over. Academia cannot replace industry and likewise. If there is any pressing problem of this symbiosis, it is more of labor attrition. More people are leaving academia for better pay. But that is not what the topic was about.

nyanpasu64 5y ago

> And then, when a corporation is involved one runs the risk of the findings disappearing behind a copyright or a patent wall. True, but universities often patent algorithmic (and other) discoveries and try to charge companies for implementing them as well.