While I agree that "science" (as defined in the article) is a very good way to make use of data, I do not think the data without the science is useless.
For one, if you collect the data now you can always apply the "science" later--as long as you have the data and can query it, you're always ready. This alone means that collecting data now could be useful even if you have no idea of what to do with it. Storage is relatively cheap, and you--or somebody else--might come up with a clever way to analyze it in the future.
Additionally, I think the author underestimates the potential of machine learning. I am by no means an expert in the field (I'm taking an AI class--that has to count for something! ;) I hope) but even the simple techniques we've covered can get some interesting information without much domain knowledge involved. As ML evolves, I suspect there will be more and more technology that can find interesting trends and relationships in data regardless of what the data is actually modelling. The article does mention ML a bit, but I think it will be much more significant in the near future.
Ultimately, figuring out what questions to ask about data and harnessing human curiosity will give you much more than just hoarding data. I just think it's possible that sufficiently abstract and generic approaches in the near future will make it easier to get similar--or perhaps orthogonal--results using techniques from ML and AI. I firmly believe that hoarding the data without doing anything to it is still much better than not collecting it at all.
Machine learning is no silver bullet. You still have to understand the algorithms and the assumptions they make. For example, if you're using a support vector machine, do you have reasonable to believe your data is linearly separable? If you're using a kernel, does the kernel make sense for your data? You can't just blindly try methods till something seems to work, because you'll end with something that works by accident on your test data but doesn't apply to the real problem you're trying to solve. Understand the data and everything else follows.
I agree with you that collecting data can be a good thing, and it can be analyzed at some later point. But in the case of online world, the lifetime of data is very small(this might not be for other industries). Trends change very rapidly, and using an old data for making inference might not be profitable if not a loss.
Actually that's exactly the opposite of true. In fact you -need- old data to validate that your mathematical model actually works by applying it to whatever historical timespan of data you are trying to derive intelligence from. At least if any sort of unit of time is applicable to any of your models.
Of course just sitting on the data doesn't do anyone any good at all..and randomly picking at stuff isn't likely to do much either unless you go in knowing what you want to find.
For one, if you collect the data now you can always apply the "science" later--as long as you have the data and can query it, you're always ready. This alone means that collecting data now could be useful even if you have no idea of what to do with it. Storage is relatively cheap, and you--or somebody else--might come up with a clever way to analyze it in the future.
Additionally, I think the author underestimates the potential of machine learning. I am by no means an expert in the field (I'm taking an AI class--that has to count for something! ;) I hope) but even the simple techniques we've covered can get some interesting information without much domain knowledge involved. As ML evolves, I suspect there will be more and more technology that can find interesting trends and relationships in data regardless of what the data is actually modelling. The article does mention ML a bit, but I think it will be much more significant in the near future.
Ultimately, figuring out what questions to ask about data and harnessing human curiosity will give you much more than just hoarding data. I just think it's possible that sufficiently abstract and generic approaches in the near future will make it easier to get similar--or perhaps orthogonal--results using techniques from ML and AI. I firmly believe that hoarding the data without doing anything to it is still much better than not collecting it at all.