I think your recommendation for NASA to develop better protocols for validating crowd-sourced solutions is vital. In your example, where 21,000 volunteers studied 100,000 images, I would imagine the volunteers used many different methods to do this task, some of which may have been applied incorrectly. Therefore, NASA would need to spend significant time and resources on quality-control on the results of this crowd sourced challenge. Moreover, maybe a few of these volunteers came up with radically different and efficient methods for detecting moving objects – how would NASA be able to identify these unique methods among the sea of submissions? So while I see how crowd sourcing can be immensely useful, I am also worried about the challenge of validating and understanding the huge amount of results it will generate.
Interesting topic – thanks for sharing. While the benefits of having an open-access platform to academic research are many, my one and only concern is that of ensuring publications are peer-reviewed. A researcher cannot publish in a respected journal before that paper has been reviewed and approved by several leading academics in the field. If people lost trust in the accuracy/credibility of the research published on an open platform, no one would use it anymore. So as one of the comments mentioned above, I think that ensuring all publications are indeed peer-reviewed is essential for any reputable open access platform. However, managing the peer-review of thousands of research articles will cost money, so how can such a platform, which does not charge people for using it, remain financially sustainable? One way is for a charitable foundation to bank-roll it, but I think a better way is to involve global governments to fund such initiatives.
Thanks for highlighting the effect machine learning can have in this field. I see the benefit of identifying new consequences or potential causes of diabetes; however, machine learning does not identify causal relationships, just correlated ones, so one could not simply state that, for example, rice consumption has led to a high incidence of diabetes. Instead, one could say that the two are correlated with one another, and this could then guide further research into the matter. So I see machine learning as a research tool to analyse complicated sets of data, not a replacement for identifying causation.
While I appreciate that the advent of automated glucose measurement devices enables the collection of useful data related to the management of diabetes, it was not exactly clear to me how machine learning will help diagnose new incidents of diabetes. Typically, I would assume that only people with diabetes or a familial history of diabetes would measure their glucose levels regularly. Therefore, I don’t see how a typical person might be identified as a high-risk candidate of having diabetes, unless he/she measures glucose levels regularly.
Mark – you adeptly summarised how 3D printing helped Shell overcome the design challenges associated with developing this pioneering technology. I can clearly see how 3D printing helped engineers visualise the solution they developed, find errors early on and thus reduce the design lead time and costs associated. I think you posed a good question: to what extent can 3D printing be scaled? For example, I have my doubts as to applicability of 3D printing to tasks such as pipework. I can definitely see how robots may be used to weld pipe segments together to create unique pipe segments, but question whether it would be efficient to 3D print such a pipe segment given the majority of that segment is of uniform cylindrical shape. If this technology evolves with time to handle such repetitive tasks efficiently, this would revolutionise the construction industry, but I still believe the iron triangle will remain. No matter your level of technological advancement, a project manager will always have a trade-off between cost, schedule and quality. Technology would just raise the bar as to what is considered acceptable standards.
I think you did a great job summarising GE’s short term and long term plans with regards to applying its preventative maintenance algorithm. However, I question whether GE as the OEM is the best suited party to develop such algorithms. There is an inherent conflict of interest since I would imagine GE derives a significant amount of its revenues from its maintenance contracts, which I know to be the case for gas turbines. Moreover, since machine learning algorithms do not utilise causal relationships (just correlated ones), I don’t see an advantage of the OEM developing such algorithms, apart from the fact that it is the only entity that can easily compile maintenance data from its customers around the world. If GE’s intention is truly to improve up-time of its machines, then it would make the raw data available on an open access platform for 3rd parties to develop their own machine learning algorithms.